• 沒有找到結果。

Adoption and Implementation of CASE Tools in Taiwan

N/A
N/A
Protected

Academic year: 2021

Share "Adoption and Implementation of CASE Tools in Taiwan"

Copied!
24
0
0

加載中.... (立即查看全文)

全文

(1)Information & Management 35 (1999) 89±112. Research. Adoption and implementation of CASE tools in Taiwan Prof. Heng-Li Yang* Department of Management Information Systems, National Cheng-Chi University, 64, Sec. 2, Chihnan Road, Mucha Dist., 116, Taipei, Taiwan Received 17 December 1997; accepted 30 August 1998. Abstract The aim of the research discussed here was to understand computer-aided software engineering (CASE) tool usage in Taiwan. Through a literature review, we developed two questionnaires ± one for general respondents, the other for teachers and CASE agents. After pre-testing, 786 questionnaires were mailed out and 226 effective responses were obtained after two follow-up letters. Factor analyses were used to condense factors from `severity of critical problems in system development', `severity of perceived problems in CASE usage', `attitude toward CASE' and `CASE implementation success determinants'. Several external variables were considered in exploring their possible in¯uence as well as the attitude and organizational features of the organizations that successfully used CASE. Path analyses were used to test an attitude model of CASE adoption and implementation success determinants. The results show that `the perceived problems in CASE tools' had no statistically signi®cant in¯uence on `attitude toward CASE' and very little in¯uence on `perceived CASE improvement for system development critical problems'. In addition, we found that `methodology use' (including the usage before CASE adoption and consistency with the methodology supported by CASE) was the only statistically signi®cant CASE implementation success determinant. Using only a `methodology use' variable could provide a way to discriminate the successful adopter from relatively unsuccessful adopter with a 75% correct classi®cation rate. # 1999 Elsevier Science B.V. All rights reserved Keywords: Computer-aided software engineering (CASE); CASE adoption; CASE success; CASE implementation. 1. Introduction The dramatic growth in the use of information technology and its dynamic environment has created a heavy demand for information systems (IS) that become available more rapidly and at reduced cost. However, system development has been recognized as a task with a high level of complexity and has been more of an art rather than a science. Computer-aided *Corresponding author. Tel.: +886-2-2938-7651; fax: +886-22939-3754; e-mail: yanh@mis.nccu.edu.tw. software engineering (CASE) provides automation of the software engineering discipline. It may be applied in different phases of the system development life cycle, and has been classi®ed into upper-, middle-, lower-plus cross-life-cycle CASE. It has also generated interest about ways to address IS development and maintenance problem. However, actual experiences with tools have exhibited more ambiguity. For example, Kemerer [12] reported that one year after introduction, 70% of the CASE tools have never been used, 25% have been used by only one group, and 5% widely used, but not to capacity. It is also suggested. 0378-7206/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved PII: S-0378-7206(98)00081-0.

(2) 90. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. that certain environmental components of a CASE implementation plan will result in a more effective and successful outcome. This paper reports on a survey of businesses, government organizations, teachers, and CASE agents to understand their attitudes toward CASE and its implementation in Taiwan. 2. Literature review In our review of literature, research papers (e.g., [4, 21, 22]) were included to determine how organizations currently perform system analysis and design. Some (e.g., [15]) identi®ed key factors to develop IS. Sumner and Ryan [27] not only identi®ed the critical success factors in each stage of IS development, but also reported users' perceived impact of CASE on these factors. They showed that the CASE users did not view CASE as having a positive impact on achieving these factors. On one hand, some research (e.g., [1, 2, 13, 16, 17, 29, 30]) concluded from empirical (experiments, surveys, or cases) or work experience that CASE tools. have the bene®ts of improving IS development productivity, quality, software reuse, documentation, maintenance, etc. However, on the other hand, some research (e.g., [3, 9, 20]) gave warning that CASE might have negative impacts, de-skilling or limiting the creativity of IS developers, or threatening job security and so inducing resistance. In a survey of the UK, Stobart, Thompson and Smith [26] identi®ed some problems with current CASE tools, e.g., poor interface or supplier support. Conger [5] suggested three requirements of an ideal CASE environment: integration; arti®cial intelligence; and multi-user support facilities. One possible reason for not widely using CASE tools may be that they are relatively expensive [11]. A number of research papers (e.g, [6, 18, 23, 24, 25]) also suggested that the key factors for adopting CASE successfully include: methodology use, pilot project use, training, consultant guide, demonstration, power-coercive strategy, top management support, software metrics use, etc. In their study, Urwiler et al. [28], predetermined two success criteria ± quality and productivity, and then found that `methodology use' and `software metrics' were signi®cant to quality;. Fig. 1. Research model of CASE attitude..

(3) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. 91. Fig. 2. Research model of CASE success.. `software metrics', `consultant use', and `training' were signi®cant to productivity. 3. Methodology The conceptual models of our research are shown in Figs. 1 and 2 (the ®rst is similar to Davis' technology acceptance model [7], but different in the attitude variable). Their signs show the hypothesized positive or negative association among the internal variables. A number of external variables are grouped as follows: personal including ± position ± time involved in the IS jobs ± specialty level of traditional system analysis and design, not applying CASE, ± specialty level of CASE tools, and ± career planning preference; organizational including ± ± ± ±. organization type, industry type, organization size, and IS growth stage;. IS department including ± history time,. ± number of IS employees, ± number of IS employees responsible for system maintenance, ± number of IS employees skilled at CASE, ± computer hardware used, ± percentage of IS budget spent on sales, and ± percentage of software development expenses spent on the IS budget; system development variables including ± ± ± ± ± ± ±. outsourcing IS or not, IS project size, type of methodology used, CASE tools use or not, number of CASE tools used, CASE usage time, and kinds of CASE used.. For simpli®cation, we might assume that external variables do not signi®cantly in¯uence the internal variables. However, this assumption needs to be tested. In this study both questionnaire and follow-up interview data collection methods were used. Through literature review, we developed a questionnaire including two versions ± one (7 pages) for `general respondents', the other (4 pages) for `teachers and CASE agents'. They were ®rst evaluated by four IS experts, and then pretested with ®fteen MIS personnel.

(4) 92. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 1 Profile of all respondents Respondent position or occupation CEO (Chief Executive Officer) CIO (Chief Information Officer) Middle manager in MIS department MIS department personnel (SA, programmer, etc.) Other department Sub-total Teacher in MIS department of universities Teacher in computer science department of universities Teacher in other departments of universities CASE vendor Sub-total Total Use of CASE tools a Currently use Used before, but abandoned Currently evaluating CASE Evaluated before, but dismissed Never considered to adopt Sub-total Wrong answer Not disclosed Total. General respondents. Teachers & CASE-industry respondents. Number 8 51 57 43 16 175. Ratio 4.6% 29.1% 32.6% 24.6% 9.1% 100%. Number. Ratio. 40 1 8 2 51. 78.4% 2.0% 15.7% 3.9% 100%. 33 25 17 21 37 133 30 12 175. 24.8% 18.8% 12.8% 15.8% 27.8% 100%. 13 10 6 9 3 41 8 2 51. 31.7% 24.4% 14.6% 22.0% 7.3% 100%. All respondents Number 8 51 57 43 16 175 40 1 8 2 51 226. Ratio 3.5% 22.6% 25.2% 19.0% 7.1% 77.4% 17.7% 0.4% 3.5% 0.9% 22.6% 100%. 46 35 23 30 40 174 52 52 226. 26.4% 20.1% 13.2% 17.2% 23% 100%. Number of CASE tools used: 72 valid `all respondents': average 2.21; 43.1% had used 1, 30.6% had used 2; among them: (1) 50 valid `general respondents': average 2.18; 44% had used 1, 28% had used 2. (2) 22 valid `teachers & CASE industry respondents': average 2.28; 40.9% had used 1, 36.4% had used 2. CASE tool usage time: 68 valid `all respondents': average 3.9 years; 20.6% had used 1 year, 80.8% 5 years, among them: (1) 49 valid `general respondents': average 4.04 years; 20.4% had used 1 year, 81.6% 5 years, (2) 19 valid `teachers & CASE industry respondents': average 3.55 years; 21.1% had used 1 year, 79% 5 years. Kinds of CASE tools used: 72 valid `all respondents', among them: 37 had used upper CASE; 15 used lower CASE; 21 used integrated CASE; 10 used cross life cycle CASE Specialty level of SA (system analysis) & SD (system design) (rating from 1 to 7): 225 valid `all respondents': average 5.25; 74.6%'s rating >3; among them: (1) 174 valid `general respondents': average 5.21; 73.6% >3; (2) 51 valid `teachers & CASE industry respondents': average 5.39; 78.4% >3 t test between (1) and (2): tˆ0.973, significant levelˆ0.333 Specialty level of CASE tools (rating from 1 to 7): 226 valid `all respondents': average 3.35; 48.3%'s rating <3; among them: (1) 175 valid `general respondents': average 3.09; 57.7% <3; (2) 51 valid `teachers & CASE industry respondents': average 4.26; 15.7% <3; 45.1%ˆ3 t test between (1) and (2): tˆ9.480, significant level <0.001 a. After cross-checking their usage numbers and usage time of CASE, we detected some respondents as `wrong answer'. Those were who answered neither `currently use' nor `used before', but reported their usage numbers or usage time..

(5) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. (including MIS graduate students, professionals, and teachers). The internal reliability (Cronbach ) was calculated for the four major sections.1 According to Guielford, our Cronbach s were high because all of them were greater than 0.70. Both versions of the questionnaire were broken into two major parts. The ®rst contained six sections including questions complied from the literature: 1. 25 questions about attitude toward CASE impacts (pre-tested Cronbach ˆ0.776); 2. 15 questions about evaluated importance of CASE key implementation success determinants (Cronbach ˆ0.737) and their corresponding achievement degrees (this half was answered only by general respondents who had already used CASE); 3. 17 questions about evaluated severity of critical problems in system development (Cronbach ˆ0.734) and the corresponding improvable degrees by CASE; 4. 13 questions about evaluated severity of perceived problems in CASE tools (Cronbach ˆ0.766); 5. one closed question and one open question about perceived future for CASE; 6. 11 listed reasons for not adopting CASE (some open spaces also provided). In Sections 1,2,3 and 4 (also the closed question of Section 5), respondents were asked to rate each question on a 5-point Likert scale. But in Section 6, they were asked to rank the reasons in order of importance. The second part contained demographic questions inquiring of the respondent about his (her) related personal, organizational, IS department, and system development variables (this part was very short for teachers and CASE agents, but longer for general respondents). Owing to space limitation, the original questionnaire (in Chinese) has not been given here, but it will be provided on request. Readers can also see those questions in the following analyses. 1 Cronbach is an index to measure the reliability (coefficient of internal consistency) of an instrument. Guielford [10] suggests that reliabilities beyond 0.70 is high, between 0.70 and 0.35 is acceptable, below 0.35 is low. Nunnally [19] argues that the accepted level of reliability depends on the purpose of the research. For early stages of research, it is suggested that reliabilities of 0.5 to 0.6 suffice, and that increasing reliabilities beyond 0.80 is probably wasteful.. 93. Table 2 Other profile of general respondents. Number. Ratio. Career planning preference Technical oriented Business or management oriented Sub-total Not disclosed Total. 48 122 170 5 175. 28.2% 71.8% 100%. Organization type Government agency Government-owned business or utility Private business Non-profit organization Other Total. 44 17 98 12 4 175. 25.1% 9.7% 56.0% 6.9% 2.3% 100%. Industry type Manufacturing Finance: banking, insurance IS software business Other service industry Other business Sub-total Not disclosed Total. 52 23 27 32 40 174 1 175. 29.9% 13.2% 15.5% 18.4% 23% 100%. Organization size Top 100 business Small/medium business Other Sub-total Not disclosed Total. 87 24 23 134 41 175. 64.9% 17.9% 17.2% 100%. IS growth stage Initial stage Expansion stage Control stage Maturity stage Total. 8 22 87 58 175. 4.6% 12.6% 49.7% 33.1% 100%. History of MIS department Within 1 year 1 to 2 years 2 to 5 years 5 to 10 years 10 to 20 years More than 20 years Sub-total Wrong answer Not disclosed Total. 4 4 27 37 79 21 172 1 2 175. 2.3% 2.3% 15.7% 21.5% 45.9% 12.2% 100%. 39 45 13 33. 30% 34.6% 10% 25.4%. Computer hardware used Mostly mainframe(s) Mostly min-computer(s) Mostly workstations Mostly PCs.

(6) 94. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 2 (Continued ) Other profile of general respondents Sub-total Wrong answer Not disclosed Total Percentage of IS budget on sales 1% 1% 5% 5% 10% 10%< Sub-total Not disclosed Total. Table 3 (Continued ) Number. Ratio. 130 44 1 175. 100%. 57 51 5 12 125 50 175. 45.6% 40.8% 4.0% 9.6% 100%. Table 3 Other profile of general respondents (continue-end). Number. Percentage of software development expenses on IS budget 1% 54 1% 5% 50 5% 10% 21 10%< 31 Sub-total 156 Not disclosed 19 Total 175 Source of IS All developed in-house Mostly developed in-house Mostly out-sourcing All out-sourcing Sub-total Not disclosed Total IS project size within recent 10 years 1±20 man-months 21±60 man-months 61±140 man-months More than 140 man-months Have all kinds (all of the proportions of the above sizes are close) Sub-total Not disclosed Total. Ratio. 34.6% 32.1% 13.5% 19.9% 100%. 67 61 31 11 170 5 175. 39.4% 35.9% 18.2% 6.5% 100%. 64 38 14 11 32. 40.3% 23.9% 8.8% 6.9% 20.1%. 159 16 175. 100%. Type of system development methodology used (1) None 16 (2) Only system flow chart 14 (3) Only data flow diagram 10 (4) Process-oriented structured 51 SA & SD only. 10.8% 9.5% 6.8% 34.5%. Other profile of general respondents (continue-end) (5) Data-oriented method only (6) Object-oriented method only (7) Others (6) Both (4) and (5) (7) Both (4) and (6) (8) Both (5) and (6) (9) Follow both (4), (5) and (6) Sub-total Wrong answer Not disclosed Total Perceived CASE tools success Very unsuccessful Unsuccessful So-so (no opinion) Successful Very successful Sub-total Wrong answer or not disclosed Total. Number. Ratio. 9 7 2 19 7 2 11 148 23 4 175. 6.1% 4.7% 1.4% 12.8% 4.7% 1.4% 7.4% 100%. 4 15 14 14 3 50 125 175. 8% 30% 28% 28% 6% 100%. Note 1: Time of respondents involved in IS jobs: averageˆ12.1, medianˆ11, maximumˆ35. Note 2: Total number of employees in organization: averageˆ1748, medianˆ600, maximumˆ25 000. Note 3: Number of employees in IS department: averageˆ44.2, medianˆ17, maximumˆ900. Note 4: Number of IS employees who were responsible for system development & maintenance: averageˆ24, medianˆ10, maximumˆ800. Note 5: Number of IS employees who were skilled at CASE: averageˆ10.2, medianˆ2, maximumˆ700. Note 6: The degree (1 to 7) of commercial CASE in¯uencing their IS outsourcing decisions: 62.5% have never been in¯uenced (choosing degree 1 to 3); 25.6% have been somewhat in¯uenced (choosing degree 5 to 7).. To maximize information quality and response rate, Dillman's `total design method' [8] was employed in designing both the content and the administrative procedures for the questionnaire. On December 4, 1996, 786 questionnaires were mailed to: Top 100 business, and government agency of®cials, people in the ®nancing industry, members of the Information Manager Association and of the Society of Information Management, and teachers of system analysis/ design or software engineering in MIS departments of universities,2 plus CASE agencies and their user 2 The teacher listing was collected from categories of universities..

(7) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. groups. Two follow-up letters were mailed on December 27 and the following January 14. If respondents provided their phone numbers, telephone interviews were later performed to clarify any unclear answers. 4. Analysis and findings Two hundred and twenty six valid questionnaires were eventually obtained from the original mailing, providing a net response rate of 29%. 4.1. Sample characteristics We asked teachers and CASE agents less demographic data. Table 1 provides the pro®le of all respondents, Tables 2 and 3 provide a pro®le of the general respondents. One question is not listed in the tables. General respondents were asked to pick one of four descriptions that best characterized the overall stage of development of IS in their organizations. The. 95. result is reported in Table 2. However, in another question, respondents gave their degree (1 to 7) of assessment of the degree of implementation of IS in their organizations; this was used to provide a measure of reliability for the `perceived IS growth stage'. Analysis of variance (ANOVA) showed that there were signi®cant differences among the scores of the four stages (Fˆ17.082, p<0.001). The Pearson coef®cient of correlation between the two questions was 0.454 at the signi®cant level less than 0.001. Therefore, there was high reliability in measuring respondents' perceived IS growth stages. 4.2. IS development and CASE usage Tables 4±10 give the results of perceived severity of critical problems in system development, perceived improvable degree by CASE to system development critical problems, perceived severity of problems in CASE tools, attitude toward CASE impacts, evaluated. Table 4 Severity of critical problems in system development Critical problems in system development. All respondents. General respondents. Teachers & CASE-agent respondents. Poor system analysis quality Poor user-analyst communication Heavy system maintenance workloads Poor system design quality Low system developer productivity Poor software quality Poor project management Poor document quality Insufficient top management support and commitment Lack of methodology to integrate various techniques and tools Lack of techniques to shorten system development life cycle Programming design and testing time span too long System analysis and design time span too long Insufficient end-user training Poor programmer-analyst communication Insufficient cost-effectiveness analysis Many current software and hardware limitations Total rating (range between 17 and 85). (1) 4.06 (2) 3.96 (3) 3.95 (4) 3.88 (5) 3.80 (6) 3.76 (7) 3.79 (8) 3.76 (8) 3.76 (10) 3.69 (11) 3.63 (12) 3.61 (13) 3.58 (14) 3.58 (15) 3.53 (16) 3.51 (17) 3.22 62.94. (1) 4 (3) 3.88 (2) 3.96 (4) 3.85 (5) 3.78 (8) 3.72 (7) 3.75 (6) 3.75 (9) 3.68 (10) 3.61 (12) 3.59 (11) 3.59 (13) 3.59 (14) 3.54 (15) 3.52 (16) 3.47 (17) 3.29 62.42. (1) 4.24 (1) 4.24 (7) 3.92 (5) 3.98 (9) 3.88 (3) 4.06 (7) 3.92 (10) 3.78 (4) 4.04 (6) 3.94 (11) 3.76 (13) 3.68 (15) 3.56 (12) 3.7 (16) 3.53 (14) 3.66 (17) 3 64.69. Note 1: The Cronbach of all respondents is 0.888. Note 2: Each cell has a pair of numbers ± i.e., (ranking #) rating #, the numbers of the parentheses in the table were the severity rankings according to their rating scores in each group (The highest score has the rank# 1). Note 3: The t test between the total rating differences between `general respondents' and `teachers & CASE-agent respondents': and tˆÿ1.618, signi®cant level ˆ0.109. Note 4: The Spearman ranking correlation of rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.816, signi®cant level <0.001..

(8) 96. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 5 Perceived improvable degree by CASE for system development critical problems Critical problems in system development. All respondents. General respondents. Teachers & CASE-agent respondents. Poor document quality Poor system design quality Low system developer productivity Heavy system maintenance workloads System analysis and design time span too long Programming design and testing time span too long Poor software quality Lack of techniques to shorten system development life cycle Poor project management Poor system analysis quality Lack of methodology to integrate various techniques and tools Poor programmer-analyst communication Poor user-analyst communication Insufficient cost-effectiveness analysis Insufficient end-user training Insufficient top management support and commitment Many current software and hardware limitations Total rating (range between 15 and 85). (1) 3.93 (8) (2) 3.61 (4) (3) 3.58 (5) (4) 3.54 (3) (5) 3.53 (13) (6) 3.52 (12) (7) 3.5 (6) (8) 3.44 (11) (9) 3.42 (7) (10) 3.41 (1) (11) 3.40 (10) (12) 3.40 (15) (13) 3.27 (2) (14) 3.08 (16) (15) 3.05 (14) (16) 2.97 (8) (17) 2.88 (17) 57.57. (1) 3.89 (6) (2) 3.56 (4) (5) 3.52 (5) (3) 3.52 (2) (7) 3.45 (13) (4) 3.52 (11) (6) 3.49 (8) (8) 3.45 (12) (12) 3.38 (7) (11) 3.38 (1) (9) 3.39 (10) (9) 3.39 (15) (13) 3.28 (3) (15) 3.14 (16) (14) 3.14 (14) (16) 3.06 (9) (17) 2.93 (17) 57.45. (1) 4.06 (10) (3) 3.77 (5) (2) 3.78 (9) (5) 3.59 (7) (4) 3.76 (15) (7) 3.53 (13) (6) 3.54 (3) (10) 3.44 (11) (7) 3.53 (7) (9) 3.51 (1) (11) 3.43 (6) (12) 3.42 (16) (13) 3.25 (1) (14) 2.90 (14) (15) 2.78 (12) (17) 2.65 (4) (16) 2.75 (17) 57.96. Note 1: Each cell has a triple of numbers ± i.e., (ranking#) rating# (ranking#), the numbers of the ®rst parentheses were the improvement rankings according to their improvement rating scores in each group (the highest score has the rank# 1); the numbers of the second parentheses were the severity rankings according to their severity rating scores in each group (also appeared in Table 4). Note 2: The t test between the total improvement rating differences between `general respondents' and `teachers & CASE-agent respondents': tˆÿ0.305, signi®cant level ˆ0.761. Note 3: The Spearman ranking correlation of improvement rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.907, signi®cant level <0.001. Note 4: The Spearman ranking correlation of between `improvement rankings' and `severity rankings' in the `general respondents' group: rsˆ0.46, signi®cant level ˆ0.063. Note 5: The Spearman ranking correlation of between `improvement rankings' and `severity rankings' in the `teachers & CASE-agent respondents' group: rsˆ0.128, signi®cant level ˆ0.624. Note 6: The Spearman ranking correlation of between `improvement rankings' and `severity rankings' in the `all respondents': rsˆ0.445, signi®cant level ˆ0.073.. importance and achievement degree of implementation success determinants, reasons of not adopting CASE, and perceived future for CASE. Except for Table 9, the total rating between `general respondents' and `teachers and CASE-agent respondents' were not signi®cantly different (by t tests); and their question rankings were signi®cantly related (by Spearman ranking correlations). For general respondents, the rankings between `perceived severity of critical problems in system development' and their `perceived improvable degree by CASE' were marginally (at ˆ0.063) signi®cantly related. Generally speaking, all respondents considered that there were critical problems in system development and had a favorable. attitude toward CASE impacts, though they also perceived some problems in CASE tools. In addition, all of them had great beliefs in CASE future improvement and user acceptance (only 2.3% perceived CASE `no future' and 16.6% perceived `little improvement on current tools'; these ratios were much smaller than the results of [26]). However, about the reasons of not adopting CASE, teachers and CASE-agents perceived `organization structures and culture', `top management support' more important, but underestimated the knowledge of general respondents about CASE and methodology. For general respondents not adopting CASE, the more important reasons were price, Chinese compatibility, and CASE quality..

(9) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. 97. Table 6 Perceived severity of problems in CASE tools Perceived problems in CASE tools. All respondents. General respondents. Teachers & CASE-agent respondents. High price Poor supplier support Lack of intelligent capabilities Poor tool integration Poor code generation Poor Chinese compatibility Poor user or reference manual Longer learning time comparing to traditional method Poor tool efficiency Poor multi-user facilities Poor user interface Poor methodology Poor generated documents Total rating (range between 13 and 65). (1) 4.14 (2) 3.81 (3) 3.75 (4) 3.66 (5) 3.58 (6) 3.57 (7) 3.49 (8) 3.46 (9) 3.37 (10) 3.46 (11) 3.26 (12) 3.16 (13) 3.14 45.93. (1) 4.10 (2) 3.81 (3) 3.70 (4) 3.62 (5) 3.53 (6) 3.53 (7) 3.48 (8) 3.44 (9) 3.4 (10) 3.34 (11) 3.26 (13) 3.14 (12) 3.19 45.63. (1) 4.26 (4) 3.84 (2) 3.92 (5) 3.78 (6) 3.74 (7) 3.72 (8) 3.52 (9) 3.5 (10) 3.28 (3) 3.88 (11) 3.26 (12) 3.2 (13) 2.98 45.88. Note 1: The Cronbach of all respondents is 0.811. Note 2: Each cell has a pair of numbers ± i.e., (ranking#) rating#, the numbers of the parentheses in the table were the severity rankings according to their rating scores in each group (The highest score has the rank# 1). Note 3: The t test between total rating differences between and `teachers & respondents': tˆÿ1.271, signi®cant level ˆ0.207. Note 4: The Spearman ranking correlation of rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.830, signi®cant level <0.001.. 4.3. Factor analyses We conducted four factor analyses to determine which variables were major factors predicting critical problems in system development, severe problems of CASE tools, attitude toward CASE impacts, and key implementation success determinants. The lower bound of factors was determined by counting the eigen-values greater than 1. In order to ®nd the upper bound, the maximum likelihood method was used to test the hypothesis: n factors are suf®cient (vs. the alternative hypothesis: more factors are needed). Within the upper and lower bounds, considering the criteria of variances explained by each factor greater than `average variable' (1/number of questions) and scree test criterion; we used the Principal Component method for extracting factors, and tried both the Orthogonal Varimax and Oblique Oblimin (in SPSS package) as the rotation methods. The ®nal factor structures are shown in Tables 11±14. Critical problems in system development There were six factors in Table 11 explaining 71.6% of the total variance. However, the. Fig. 3. Model of attitude toward CASE (all respondents Nˆ174)..

(10) (2) 4.04 (3) 4.04 (4) 3.95 (5) 3.92 (6) 3.80 (7) 3.77 (9) 3.71 (11)3.69 (10) 3.71 (8) 3.73 (12) 3.69 (13) 3.56 (14) 3.51 (15) 3.49 (17) 3.34 (20) 3.29 (16) 3.35 (19) 3.31 (21) 3.26 (18) 3.31 (22) 3.19 (23) 3.04 (24) 3.02 (25) 2.91 88.71. (2) 4.07 (3) 4.06 (4) 4.02 (5) 3.95 (6) 3.86 (7) 3.79 (8) 3.79 (9) 3.73 (10) 3.72 (11) 3.71 (12) 3.69 (13) 3.56 (14) 3.56 (15) 3.53 (16) 3.37 (17)3.43 (18) 3.34 (19) 3.28 (20) 3.27 (21) 3.27 (22) 3.19 (23) 3.14 (24) 2.96 (25) 2.84 89.06. (20) 3.28 (21) 3.2 (19) 3.33 (23) 3.1 (22) 3.18 (17) 3.49 (24) 2.74 (25) 2.62 90.28. (3) 4.16 (1) 4.26 (6) 4.04 (5) 4.08 (9) 3.84 (7) 4.02 (8) 3.86 (10) 3.74 (13) 3.64 (12) 3.68 (15) 3.57 (11) 3.71 (14) 3.64 (18) 3.48 (16) 3.53. (2) 4.18. (4) 4.14. Teachers and CASEagent respondents. Note 1: The Cronbach of all respondents is 0.866 Note 2: Each cell has a pair of numbers ± i.e., (ranking #) rating#, the numbers of the parentheses in the table were the rankings according to their rating scores in each group (The highest score has the rank# 1). Note 3: The t test between the total rating differences between `general respondents' and `teachers & CASE-agent respondents': tˆÿ0.938, signi®cant level ˆ0.35. Note 4: The Spearman ranking correlation of impact rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.929, signi®cant level <0.001.. (1) 4.08. (1) 4.10. The developed system documents would closely follow project documentation standards and formats The outputs of all system development stages would be more consistent each other, and be also consistent with final documentation Easier to develop prototypes Easier to draw graphs for documentation Increase system development productivity ± reduce time, personnel, and cost The developed system documents is easier to understand and help communication The developed system is easier to maintain Would not threaten job security of IS developers Give IS developers valuable skills of using CASE Improve system development methodology The developed system would increase software reuse More likely to perform design alternatives before programming The developed system is more reliable Would not induce resistance of IS developers Automate the tedious work in system development Process of system development is more enjoyable to IS developers Make top managers more satisfactory and improve awareness of the efforts of developing new systems The developed system would more meet user requirements The developed system would make IS developers more satisfactory Make organizational outside customers more satisfactory Obviate the need of rigorous project management Make IS user departments (e.g., finance, marketing) more satisfactory Would not limit the creativity of IS developers to find solutions Reduce the dependence on system designers Reduce the dependence on system analysts Total rating (range between 25 and 125). General respondents. All respondents. Attitude toward CASE impact. Table 7 Attitude toward CASE impacts. 98 Prof. H.-L. Yang / Information & Management 35 (1999) 89±112.

(11) (2)4.33 (3)4.23 (5) 4.16 (7) 4.08 (6) 4.11 (4) 4.17 (8) 4 (9) 3.89 (11)3.82 (10)3.84 (13) 3.75 (12) 3.77 (14) 3.27 (15) 2.94 58.71. (2) 4.34 (3) 4.26 (4) 4.17 (5) 4.15 (6) 4.13 (7) 4.13 (8) 4.06 (9) 3.94 (10) 3.9 (11) 3.87 (12) 3.83 (13) 3.76 (14) 3.36 (15) 2.92 59.23. (15) 2.86 60.9. (14) 3.66. (13) 3.72. (9) 4.12. (10) 4.1 (8) 4.16 (12) 3.98. (7) 4.20 (11) 4.02 (5) 4.26. (6) 4.22 (2) 4.4. (1) 4.42 (4) 4.38. (2) 4.4. Importance scores of teachers and CASE-agent respondents. (15) 2.28 (15) 47.54. (10) 3.05 (14). (14) 2.45 (10). (8) 3.12 (11). (11) 3.02 (7) (9) 3.10 (13) (13) 2.79 (12). (7) 3.26 (3) (4) 3.5(2) (12) 2.95 (9). (5) 3.41 (5) (5) 3.41 (7). (3) 3.54 (6) (1)3.62 (3). (2) 3.55 (1). Achievement scores of general respondents who have used CASE (Nˆ43). Note 1: The Cronbach of importance scores of all respondents is 0.846. Note 2: Each cell of the 2nd±4th columns has a pair of numbers ± i.e., (ranking#) rating#, the numbers of the parentheses in the table were the importance rankings according to their importance rating scores in each group (the highest score has the rank# 1). Note 3: Each cell of the 5th column has a triple of numbers ± i.e., (ranking#) rating# (ranking#), the numbers of the ®rst parentheses were the achievement rankings according to the actual achievement rating scores of those `general respondents' who had used CASE; the numbers of the second parentheses were their corresponding importance rankings (Nˆ43). Note 4: The t test between the total importance rating differences between `general respondents' and `teachers & CASE-agent respondents': tˆÿ1.9705, signi®cant level ˆ0.052. Note 5: The Spearman ranking correlation of importance rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.794, signi®cant level <0.001. Note 6: The Spearman ranking correlation of importance rankings between `general respondents' and `general respondents who have used CASE': rsˆ0.844, signi®cant level <0.001. Note 7: The Spearman ranking correlation between `evaluated importance rankings' and `actual achievement rankings' in the group of `general respondents who have used CASE' (Nˆ43): rsˆ0.792, signi®cant level <0.001.. (1) 4.34. (1) 4.35. Provide formal training on system development techniques required to use CASE Introduce CASE technology via at least one pilot project Facilitate knowledge sharing and open communication among project team members Enforce a system development methodology before using CASE Secure top management support and commitment on budgets and necessary organizational changes Retain experienced consultants to guide project using CASE Enhance system analysis and design abilities of system developers Periodically evaluate the CASE impacts on system development productivity and quality Help career path planning of IS developers Demonstrate and promote the benefits of using CASE Periodically evaluate the CASE impacts on organizational culture and structure, and make adjustments if necessary Use the same system development methodology before and after implementing CASE Use standard software metrics to measure the effectiveness of system development Guarantee job security of IS developers to lower the potential resistance Apply power-coercive strategy to enforce CASE Total rating (range between 15 and 75). Importance scores of general respondents (Nˆ174). Importance scores of all respondents. CASE Implementation success determinants. Table 8 Evaluated importance and achievement degree of implementation success determinants. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112 99.

(12) 100. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 9 Reasons of not adopting CASE tools Reasons of not adopting CASE. All respondents. General respondents. Teachers & CASE-agent respondents. High tool price Do not understand the methodology supported by CASE Organizational culture or structure does not fit with CASE tool Poor CASE quality Poor Chinese compatibility Not aware of CASE Cannot help system development productivity Satisfied with current system development methodology Top management do not support Poor quality of generated systems Persistent resistance from IS developers. 1 2 3 4 5 6 7 8 9 10 11. 1 5 6 3 2 7 4 8 10 9 11. 3 1 2 6 8 5 9 7 4 11 10. (4.75) (5.0) (5.46) (5.56) (5.63) (5.95) (5.98) (6.19) (6.53) (6.92) (7.10). (4.4) (5.63) (5.76 (5.4) (5.21) (6.11) (5.41) (6.24) (7.10) (6.91) (7.32). (5.26) (4.09) (5.02) (5.8) (6.22) (5.71) (6.78) (6.02) (5.70) (6.92) (6.79). Note 1: Each cell has a pair of numbers ± i.e., ranking #(scores #), the numbers of the parentheses in the table were the scores corresponding to their average rankings; the top #1 reason giving score 1, #2 score 2, etc. If a respondent only gave several top reasons (e.g., giving top 3 reasons), other reasons would be treated the average score of his (her) unanswered reasons (e.g., 4 and 11 giving the average of 7.5). Note 2: The Spearman ranking correlation of rankings between `general respondents' and `teachers & CASE-agent respondents': rsˆ0.418, signi®cant level ˆ0.201. Note 3: In the free space provided in the questionnaire, only two respondents gave two reasons which were not related to the above, but related to the external environments of an organization: `IT changed too fast' and `CASE marketing prevalence rate was low'. An anonymous reviewer mentioned a reason `worrying that the developed system would be dependent on the CASE tool and trained users' could be classifed into `orgainzational culture or structure does not ®t the CASE tool' as above. Note 4: The Spearman ranking correlation of rankings between `IS developed in-house' and `purchasing packages or outsourcing IS' respondents: rsˆ0.864, signi®cant level ˆ0.001. So, the opinions of both groups of respondents are similar. Table 10 Perceived future for CASE tools Future for CASE tools. (1) No future (2) Little improvement on current tools (3) Improve substantially and the market acceptance of tools increases (4) Become accepted by many system developers as preferred methods of system development (5) Totally alters the way that all software is developed and maintained Sub-total Average Not answered Total. All respondents. General respondents. Teachers & CASE-agent respondents. Number. Ratio. Number. Number. Ratio. 2 27 88. 1.2% 16.2% 52.7%. 3 9 19. 6% 18% 38%. 5 36 107. 2.3% 16.6% 49.3%. 31. 18.6%. 14. 28%. 45. 20.7%. 19. 11.4%. 5. 10%. 24. 11.1%. 167 3.23 8 175. 100%. 50 3.18 51 51. Ratio. 100%. 217 3.22 9 226. 100%. Note: If giving the ®ve categories the rating of 1 to 5, the t test of future scores between `general respondents' and `teachers & CASE-agent respondents' would be: tˆ0.292, signi®cant level ˆ0.772.. Cronbach of the sixth factor3 was found to be too low. Therefore, we could only obtain ®ve factors 3 This sixth factor already existed even if we tried 5-factor structure.. plus two variables: poor system quality and top management support (factor 1), long development time and heavy maintenance (factor 2), lack of techniques and methodologies (factor 3), poor communication (factor 4), poor management (factor 5),.

(13) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. 101. Table 11 Factor matrix (varimax-rotated) of critical problems in system development Critical problems in system development (their rankings in the parenthesis). Communalities Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6. Poor system analysis quality (1) Poor system design quality (4) Poor software quality (6) Insufficient top management support and commitment (8) System analysis and design time span too long (13) Programming design and testing time span too long (12) Heavy system maintenance workloads (3) Lack of techniques to shorten system development life cycle (11) Lack of methodology to integrate various techniques and tools (10) Poor user-analyst communication (2) Poor programmer-analyst communication (15) Insufficient end-user training (14) Low system developer productivity (5) Poor project management (7) Insufficient cost-effectiveness analysis (16) Poor document quality (8) Many current software and hardware limitations (17) Eigen-value Variance explained by each factor Cronbach of each factor Average rating of the severity in each factor. 0.770 0.793 0.720 0.506 0.724 0.746 0.714 0.747. 0.798 0.834 0.793 0.526. 0.8 0.824 0.634. 0.709 0.798 0.716 0.694 0.604 0.718 0.698 0.815 0.703. poor documents, and current software and hardware limits. Severe problems of CASE tools Five factors in Table 12 explaining 66.1% of the total variances: poor CASE quality and support (factor 1), poor interface (factor 2), poor technical characteristics (factor 3), cost problems (factor 4, including tangible monetary expenditures and intangible learning investment), and poor advanced capacity (factor 5). Attitude toward CASE impacts Eight factors in Table 13 explaining 67.2% of the total variances: increase satisfaction (factor 1), enhance quality (factor 2), reduce dependence and rigor (factor 3), increase productivity (factor 4), facilitate documentation (factor 5), assist design (factor 6), not inducing unemployment and resistance (factor 7), and skill impacts (factor 8).. 0.794 0.808 0.776 0.603 0.622. 6.188 36.4 0.823 3.87. 1.860 10.9 0.733 3.72. 1.410 8.3 0.748 3.66. 1.002 5.9 0.768 3.69. 0.643 0.581 0.691 0.899 5.3 0.684 3.7. ÿ0.493 0.752 0.816 4.8 0.145 3.49. Key implementation success determinants Five factors in Table 14 explaining 64.3% of the total variances: communication and training (factor 1), evaluation and job security (factor 2), methodology use (factor 3), power-coercive strategy (factor 4), and software metrics (factor 5). 4.4. Features of organizations that used CASE successfully As an exploratory study, we also investigated the organizational characteristics of CASE adopters and successful users. Tables 15 and 16 list the variables marginally signi®cant on 2 tests. Although satisfying the Morrison's [14] criteria, 4 their classi®cation rates are not high. Among them, the variable having the highest classi®cation power is `number of IS 4 Morrison suggests to compute a ratio as a criterion of the chance proportion correctly classified. For example, in Table 15, the proportional chance of adopting CASE owing to the long history of IS department is (57/131)2‡(1ÿ(57/131))2ˆ0.508..

(14) 102. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 12 Factor matrix (varimax-rotated) of perceived problems in CASE tools Perceived problems in CASE tools (their rankings in the parenthesis). Communalities. Poor generated documents (13) Poor user or reference manual (7) Poor supplier support (2) Poor methodology (12) Poor user interface (11) Poor Chinese compatibility (6) Poor tool integration (4) Poor code generation (5) Poor tool efficiency (9) Longer learning time comparing to traditional method (8) High price (1) Poor multi-user facilities (10) Lack of intelligent capabilities (3) Eigen-value Variance explained by each factor Cronbach of each factor Average rating of the severity in each factor. 0.566 0.693 0.452 0.586 0.759 0.708 0.683 0.768 0.641 0.685 0.625 0.642 0.779. Factor 1 0.640 0.781 0.484 0.557. 4.032 31.0 0.632 3.40. Factor 2. 0.801 0.785. 1.278 9.8 0.689 3.42. Fig. 4. Model of attitude toward CASE (all respondents Nˆ135).. Factor 3. 0.737 0.810 0.609. 1.166 9.0 0.673 3.54. Factor 4. 0.762 0.735 1.105 8.5 0.508 3.8. Factor 5. 0.590 0.856 1.007 7.7 0.573 3.61.

(15) 0.654 0.729 0.719 0.788 0.711 0.543 0.480. 0.682. 0.665 0.672 0.623. 0.876 0.886 0.602 0.671 0.635. 0.693 0.733 0.534 0.514. 6.968 27.9 0.846 3.30. 2.385 9.5 0.776 3.78. 0.661 0.801 0.525 0.585. 1.606 6.4 0.763 3.02. 0.913 0.913 0.476. 1.470 5.9 0.701 3.62. 0.725. 0.764 0.631. 1.218 4.9 0.686 3.99. 0.722. 0.738 0.621. 1.170 4.7 0.530 3.87. 0.627 0.765. 1.066 4.3 0.587 3.67. 0.734 0.854. 0.681 0.543 0.483 0.906 3.6 0.435 3.53. 0.515 0.594 0.797 0.806 0.667. The developed system would more meet user requirements (18) The developed system would make IS developers more satisfactory (19) Make IS user departments (e.g., finance, marketing) more satisfactory (22) Make organizational outside customers more satisfactory (20) Make top managers more satisfactory and improve awareness of the efforts of developing new systems (17) The developed system is more reliable (13) The developed system is easier to maintain (7) The developed system would increase software reuse (11) The outputs of all system development stages would be more consistent each other, and be also consistent with final documentation (2) Reduce the dependence on system analysts (25) Reduce the dependence on system designers (24) Obviate the need of rigorous project management (21) Automate the tedious work in system development (15) Increase system development productivity ± reduce time, personnel, and cost (5) Process of system development is more enjoyable to IS developers (16) Easier to draw graphs for documentation (4) The developed system documents is easier to understand and help communication (6) The developed system documents would closely follow project documentation standards and formats (1) Easier to develop prototypes (3) More likely to perform design alternatives before programming (12) Would not threaten job security of IS developers (8) Would not induce resistance of IS developers (14) Would not limit the creativity of IS developers to find solutions (23) Give IS developers valuable skills of using CASE (9) Improve system development methodology (10) Eigen-value Variance explained by each factor Cronbach of each factor Average rating of the impacts in each factor. 0.411 0.511 0.840 0.877 0.756. Communalities Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 Factor 7 Factor 8. Attitude toward CASE Impact (their rankings in the parenthesis). Table 13 Factor matrix (varimax-rotated) of attitude toward CASE impact. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112 103.

(16) 104. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Table 14 Structure matrix (oblimin-rotated) of CASE key implementation success determinants CASE implementation success determinants (their rankings in the parenthesis). Communalities. Provide formal training on system development techniques required to use CASE (1) Retain experienced consultants to guide project using CASE (6) Introduce CASE technology via at least one pilot project (2) Secure top management support and commitment on budgets and necessary organizational changes (5) Facilitate knowledge sharing and open communication among project team members (3) Enhance system analysis and design abilities of system developers (7) Demonstrate and promote the benefits of using CASE (10) Guarantee job security of IS developers to lower the potential resistance (14) Help career path planning of IS developers (9) Periodically evaluate the CASE impacts on organizational culture and structure, and make adjustments if necessary (11) Periodically evaluate the CASE impacts on system development productivity and quality (8) Enforce a system development methodology before using CASE (4) Use the same system development methodology before and after implementing CASE (12) Apply power-coercive strategy to enforce CASE (15) Use standard software metrics to measure the effectiveness of system development (13) Eigen-value Variance explained by each factor Cronbach of each factor Average rating of the importance in each factor. 0.628. 0.770. 0.691. 0.810. 0.457 0.616. 0.468 0.614. 0.684. 0.790. 0.440. 0.580. employees skilled at CASE'; this distinguishes adopters from non-adopters. However, this might be the result of adopting CASE, not its cause. This variable and `use integratedly' could also distinguish successful CASE users from unsuccessful with the higher classi®cation power. 4.5. Analysis of CASE attitude model By applying total scores of internal variables and not considering the external variables in our model of CASE attitude, we obtained Fig. 3. In the path analyses, we only considered the multiple (or simple) regression equations that were signi®cant at ˆ0.05 (by ANOVA F tests), and also only drew their standardized regression coef®cient (i.e.,

(17) ) links that were signi®cant at 0.05 (by t tests).. Factor 1. Factor 2. 0.429 0.680. 0.455 0.637. 0.591 0.723. 0.701 0.843. 0.793. 0.808. Factor 3. 0.740. 0.795. 0.763. 0.833. 0.735 0.678. Factor 4. 0.846 5.072 33.8 0.789 4.23. 1.405 9.4 0.761 3.83. 1.166 7.8 0.615 4.00. 1.074 7.2 N/A 2.92. Factor 5. 0.746 0.936 6.2 N/A 3.76. The `perceived severity of problems in CASE tools' had no statistically signi®cant in¯uence on `attitude toward CASE'. Though the attitude variable of our model was different from Davis' technology acceptance model (he considered the in¯uence of `perceived usefulness' and `perceived ease of use' on `attitude towards use'), this result was still surprising. Next, we considered external variables. Originally, there were 23 in the model. In order to understand their possible in¯uence on internal variables and the underlying factors, we conducted hundreds of ANOVA F tests. Owing to space limitation, we will not report the detailed results here. Also, we were only interested in those external variables in¯uencing ®nal attitude. There were three external variables signi®cantly in¯uencing attitude of all respondents: `degree of tradi-.

(18) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. 105. Table 15 Features of organizations that have adopted CASE Features of organization. Currently use or used before. Have not adopted CASE. Subtotal. IS department has established within 10 years IS department has established more than 10 years Subtotal1. 16 (12.2%) 41 (31.3%) 57 (43.5%). 37 (28.2%) 37 (28.2%) 74 (56.5%). 53 (40.5%) 78 (59.5%) 131 (100%). Many IS employees (> median 18) Few IS employees (median 18) Subtotal2. 20 (16.3%) 42 (34.1%) 62 (50.4%). 31 (25.2%) 30 (24.4%) 61 (49.6%). 51 (41.5%) 72 (58.5%) 123 (100%). Many IS employees skilled at CASE (> median 2) Few IS employees skilled at CASE (median 2) Subtotal3. 14 (12.7%) 36 (32.7%) 50 (45.5%). 51 (46.4%) 9 (8.2%) 60 (54.5%). 65 (59.1%) 45 (40.9%) 110 (100%). Note 1: Pearson 2ˆ6.428, dfˆ1, ˆ0.011 (Classi®cation correct rateˆ59.5%> Morrison's (1969) proportional chance criterion 50.8%). Continuity correctionˆ5.55, dfˆ1, ˆ0.018. No cell had expected count less than 5. Note 2: Pearson 2ˆ4.365, dfˆ1, ˆ0.037 (Classi®cation correct rateˆ59.4%> Morrison's proportional chance criterion 50%). Continuity correctionˆ3.633, dfˆ1, ˆ0.057. No cell had expected count less than 5. Note 3: Pearson 2ˆ36.655, dfˆ1, <0.001 (Classi®cation correct rateˆ79.1%> Morrison's proportional chance criterion 51.7%). Continuity Correctionˆ34.335, dfˆ1, <0.001. No cell had expected count less than 5. Table 16 Features of organizations that used CASE successfully Features of organization. More successfully use CASE. Less successfully use CASE. Subtotal. Immature IS development stage Mature IS development stage Subtotal1. 13 (36.1%) 6 (16.7%) 19 (52.8%). 7 (19.4%) 10 (27.8%) 17 (47.2%). 20 (55.8%) 16 (47.2%) 36 (100%). Many IS employees are skilled at CASE (> median 5) Few IS employees are skilled at CASE ( median 5) Subtotal2. 12 (40%) 4 (13.3%) 16 (53.3%). 6 (20%) 8 (26.7%) 14 (46.7%). 18 (60%) 12 (40%) 30 (100%). Most of IS applications was developed in house Most of IS applications was outsourced or purchased Subtotal3. 14 (40%) 5 (14.3%) 19 (54.3%). 16 (45.7%) 0 (0%) 16 (45.7%). 30 (85.7%) 5 (14.3%) 35 (100%). No methodology or only knows to draw system flow chart or data flow diagram Follows process, data or object-oriented methodology Subtotal4. 6 (18.8%). 1 (3.1%). 7 (21.9%). 11 (34.4%) 17 (53.1%). 14 (43.8%) 15 (46.9%). 25 (78.1%) 32 (100%). Has used CASE integratedly Doesn't use CASE integratedly Subtotal5. 4 (11.1%) 15 (41.7%) 19 (52.8%). 9 (25.0%) 8 (22.2%) 17 (47.2%). 13 (36.1%) 23 (63.9%) 36 (100%). Note 1: Pearson 2ˆ2.697, dfˆ1, ˆ0.101 (classi®cation correct rateˆ63.9%> Morrison's proportional chance criterion 50.2%). No cell had expected count less than 5. Note 2: Pearson 2ˆ3.214, dfˆ1, ˆ0.073 (classi®cation correct rateˆ66.7%> Morrison's proportional chance criterion 50.2%). No cell had expected count less than 5. Note 3: Pearson 2ˆ4.912, dfˆ1, ˆ0.027 (classi®cation correct rateˆ60%> Morrison's proportional chance criterion 50.4%). Two cells had expected count less than 5. So, ran Fisher's exact test: exact probabilityˆ0.049 (two-tailed), 0.036 (one-tailed) Note 4: Pearson 2ˆ3.821, dfˆ1, ˆ0.051 (classi®cation correct rateˆ62.5%> Morrison's proportional chance criterion 50.2%). Two cells had expected count less than 5. So, ran Fisher's exact test: exact probabilityˆ0.088 (two-tailed), 0.061 (one-tailed). Note 5: Pearson 2ˆ3.955, dfˆ1, ˆ0.047 (classi®cation correct rateˆ66.7%> Morrison's proportional chance criterion 50.2%). No cell had expected count less than 5. `Use CASE integrated' means that the respondent used an integrated CASE (i.e., the CASE covering upper and lower CASE capability); or used at least two CASES ± one upper CASE and another lower CASE.

(19) 106. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Fig. 5. Model of attitude toward CASE (general respondents Nˆ100).. tional SA (system analysis) and SD (system design) skills', `degree of CASE skills', and `used CASE before'. Among these three, `degree of traditional SA and SD skills' had no signi®cant in¯uence on attitude of general respondents, however, `being an IS software company' had. The attitude of teachers and CASE agent respondents showed four external variables: `degree of traditional SA and SD skills', `being a MIS department teacher', `the number of CASE used', and `integratedly used CASE'. Considering external variables, `Perceived severity of problems in CASE tools' had still no signi®cant in¯uence on `attitude toward CASE'. In Fig. 4, `degree of CASE skills' had positive in¯uence on. all respondents' `perceived improvable degree by CASE in system development', and `used CASE before' had on their attitude. In Fig. 5, `degree of CASE skills' had similar positive in¯uence on general respondent's `perceived improvable degree by CASE', but `used CASE before' had negative in¯uence. It seems that the experience might disappoint users. To teachers and CASE agent respondents, it is not valuable to include all of the four external variables in the regression equation of their attitude. After trying several combinations, we found that the only signi®cant combination of two external variables was `degree of traditional SA and SD skills', and `being MIS department teacher'. In Fig. 6, it was.

(20) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. 107. Fig. 6. Model of attitude toward CASE (teachers & CASE agent respondents Nˆ44).. found that respondents who were less skilled in traditional SA and SD perceived more severity of problems in CASE tools. Furthermore, we considered the underlying factors of all respondents: no any factor of `perceived severity of problems in CASE tools' had signi®cant in¯uence on `attitude toward CASE', as shown in Fig. 7. However, three factors had signi®cant in¯uence on some factors of `perceived improvable degree by CASE in system development'. Possibly perceived `cost problem' had negative in¯uence on degree of CASE `improving current software and hardware limits'; and `poor CASE quality support' had negative effect on `improving development time and heavy maintenance'. However, it might be hard to explain that the more that respondents perceived `poor CASE quality and support', the more they believed that CASE tools were `improving current software and hardware limits'. We assumed that respondents might have so much expectation of the value of CASE usage that the more. problems they found in the use, the more improvement they wanted. In addition, respondents who were less skilled in traditional SA and SD felt more `lack of system development techniques and methodologies'. The results should not be considered universal because the coef®cients of determination (R2) of the regression equations were not high (about 0.3) though signi®cant. We felt that the external variable `used CASE before' might have little effect. Since there were ®ve states of this variable, we split the respondents into ®ve groups for further analysis. Because of the small size of the split samples, there were more insigni®cant regression equations. However, two points were worth noting. First, `perceived severity of problems in CASE tools' still had no signi®cant in¯uence on `attitude toward CASE' in any split group. Second, some equations had much higher R2; for example, to the attitude of all respondents who used CASE before, but abandoned, R2 of the regression coef®cient was 0.6. In addition, its stan-.

(21) 108. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Fig. 7. Detailed model of attitude toward CASE (all respondents Nˆ135).. dardized coef®cient of `perceived improvable degree by CASE in system development' was high (0.74). These con®rmed our conjecture of CASE usage effect. 4.6. Analysis of CASE success model Fig. 8 shows that the respondent's evaluated importance of CASE key implementation success factors could in¯uence his (her) degree of organizational achievement; this would further in¯uence the level of CASE success and perceived CASE impacts. It seems that, in general, the perceived CASE impacts might be good surrogates of success level. The detailed model is shown in Fig. 9. Only three determinants were evaluated to be of suf®cient importance to carry into their effect. Though the remaining two. determinants ± `methodology use' and `software metrics' were not signi®cantly important, the implementation of the ®rst could in¯uence `increasing satisfaction' (user, IS developers, customer, and top manager's satisfaction); and the second could in¯uence `reducing dependence (on analysts and designers) and project management rigor'. But only methodology use would in¯uence success level. In such a situation, the onlygoodsurrogateofsuccesslevelmightbe`increasing satisfaction'. The R2 of these multiple regression equations were between 0.4 and 0.5. Next, we considered ®ve external variables as shown in Fig. 10. `Evaluated importance of methodology use' in¯uenced its achievement, and `methodology use' became the only signi®cant determinant of either success level or its surrogates ± CASE.

(22) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Fig. 8. Model of CASE success (excluding external variables Nˆ35).. impacts. Interestingly, both `following formal methodology' and `methodology use' were important to success level. We should notice from Table 14 that there were two sub-determinants of `methodology use' ± `enforcing methodology before CASE' and `using the same methodology after implementation'. The variable `following formal methodology' in the questionnaire only captured the current situation (i.e., after implementing CASE) of the respondent's associated organization. Certainly there was some correlation (Pearson coef®cientˆ0.396) between this variable and methodology use. However, both signi®cance to success level might imply that even if an organization had no methodology before CASE, it could succeed after introducing CASE. It was also apparent that both general respondents and teachers and CASE agent respondents evaluated the importance of `enforcing methodology before CASE' rela-. 109. tively highly, and so were the corresponding achievement of the general respondents who have used CASE. However, `using the same methodology after implementation' was not. In addition, though the signi®cance (0.119) of the standardized coef®cient (

(23) ˆ0.52) of `communication and training' in the regression equation for success level was still not high, its signi®cance was only next to `methodology use' among the ®ve determinants. On the other hand, the evaluated importance of `communication and training' was the highest in Table 14. These might imply that using a new methodology after implementation must accompany with communication and training. Compared to the study of Urwiler et al., this research considered more variables and found `methodology use' to be the only important one. `Communication and training' (including consultation) was marginal and complementary to `methodology use'. As an exploratory study, we tried to use discriminant analysis to classify CASE success. We classi®ed ®ve categories of success from the corresponding success levels (1 to 5), and also two categories of success by grouping 1 and 2 into `unsuccessful' and 4 and 5 into `successful'. Without considering external variables, only `methodology use' could classify ®ve categories with correct rate 42.9%, and classify two categories with correct rate 75%. Using `methodology use' with `following formal methodology', the classi®cation correct rate would be 53.3% for ®ve categories and 81% for two categories. If we considered all ®ve determinants and all ®ve external variables, the correct rate was 76% for ®ve categories and 100% for two categories. 5. Conclusions This study, based on previous work in CASE adoption and implementation, examined CASE tool usage in Taiwan. The results of data analyses revealed that generally speaking, all respondents perceived critical problems in system development and had favorable attitude toward CASE impacts and CASE future though they also perceived some problems in CASE tools. The favorable bias might be so strong that `perceived severity of problems in CASE tools' had no signi®cant in¯uence on `attitude toward CASE' and very little in¯uence on `perceived CASE improve-.

(24) 110. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Fig. 9. Detailed model of CASE success (excluding external variables Nˆ35).. Fig. 10. Detailed model of CASE success (including external variables Nˆ25)..

(25) Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. ment for system development critical problems'. Successful implementation of CASE tools appeared signi®cantly to depend only on the `methodology usage'. In addition, using only the `methodology use' variable could predict a relatively successful adopter from relatively unsuccessful adopter with 75% of the time. However, `methodology use' was not the highest important factor evaluated by respondents. This research also provided some possible characteristics of organizations that used CASE successfully and some reasons for not adopting CASE. Acknowledgements This research is sponsored by National Science Council, Taiwan, Project # NSC 86-2416-H-004003. The author wishes to thank the chief editor, Dr. E.H. Sibley, and anonymous reviewers for their comments. References [1] R.D. Banker, R.J. Kauffman, Reuse and productivity in integrated computer-aided software engineering: an empirical study, MIS Quarterly 15(3), 1991, pp. 375±401. [2] D.L. Burkhard, Implementing case tools, Journal of System Management 4(5), 1989, pp. 20±25. [3] M. Chen, J.F. Nunamaker, E.S. Weber, Computer-aided software engineering: present status and future directions, DATA BASE, Spring 1989, pp. 7±13. [4] M.A. Colter, A comparative examination of systems analysis techniques, MIS Quarterly 8(1), 1984, pp. 51±67. [5] S. Conger, The New Software Engineering, Wadsworth Publishing Co., Belmont, California, 1994. [6] G.F. Corbitt, R.J. Norman, CASE implementation: accounting for the human element, Information and Software Technology 33(9), 1991, pp. 637±640. [7] F.D. Davis, A Technology Acceptance Model for Empirically Testing New End-User Information Systems: Theory and Results, Doctoral Dissertation, Sloan School of Management, Massachusetts Institute of Technology, 1986. [8] D.A. Dillman, Mail and Telephone Surveys ± The Total Design Method, Wiley Interscience, New York, 1978. [9] M.L. Gibson, C.A. Snyder, R.K. Rainer, CASE: clarifying common misconceptions, Journal of System Management 40(5), 1989, pp. 12±19. [10] J.P. Guielford, Fundamental Statistics in Psychology and Education, 4th ed., McGraw-Hill, New York, 1965. [11] J. Iivari, Why are CASE tools not used, Communication of the ACM 39(10), 1996, pp. 94±103. [12] C.F. Kemerer, How the learning curve affects CASE tool adoption, IEEE Software 9(3), 1992, pp. 23±28.. 111. [13] G.C. Low, D.R. Jeffery, Software development productivity and back-end CASE tools, Information and Software Technology 33(9), 1991, pp. 616±621. [14] D.G. Morrison, On interpretation of discriminant analysis, Journal of Marketing Research 6(2), 1969, pp. 159±163. [15] C.R. Necco, C.L. Gordon, N.W. Tsai, Systems analysis and design: current practices, MIS Quarterly 11(4), 1987, pp. 461±476. [16] C.R. Necco, N.W. Tsai, K.W. Holgeson, Current usage of CASE software, Journal of Systems Management 40(5), 1989, pp. 6±12. [17] R.J. Norman, J.F. Nunamaker, CASE productivity perceptions of software engineering professions, Communication of the ACM 32(9), 1989, pp. 1102±1108. [18] R.J. Norman, G.F. Corbitt, M.C. Butler, D.D. McElroy, CASE technology transfer: a case study of unsuccessful change, Journal of System Management 40(5), 1989, pp. 33±37. [19] J.C. Nunnally, Psychometric Theory, McGraw-Hill, New York, 1967. [20] W.J. Orlikowski, CASE tools as organizational change: investigation incremental and radical changes in systems development, MIS Quarterly 17(3), 1993, pp. 309±340. [21] P. Palvia, J.T. Nosek, System life cycle methods: An evaluation of their applicability, Proceedings Of the 23rd Annual Hawaii International Conference on System Sciences, 1990, pp. 141±149. [22] P. Palvia, J.T. Nosek, A ®eld examination of system life cycle techniques and methodologies, Information and Management 25(2), 1993, pp. 73±84. [23] G. Premkumar, M. Potter, Adaption of computer aided software engineering (CASE) technology: An innovation adaption perspective, Data Base Advances, 26(2 and 3), May/August 1995, pp. 105±123. [24] A. Rai, R. Patnayakuni, A structural model for CASE adoption behavior, Journal of Management Information Systems 13(2), 1996, pp. 205±234. [25] J.A. Senn, J.L. Wynekoop, The other side of CASE implementation best practice for success, Information Systems Management 12(4), 1995, pp. 7±14. [26] S.C. Stobart, J.B. Thompson, P. Smith, Use, problems, bene®ts and future direction of computer-aided software engineering in United Kingdom, Information and Software Technology 33(9), 1991, pp. 629±636. [27] M. Sumner, T. Ryan, The impact of CASE: Can it achieve critical success factors, Journal of Systems Management, June 1994, pp. 16±21. [28] R. Urwiler, N.K. Ramarapu, R.B. Wilkes, M.N. Frolick, Computer-aided software engineering: The determinants of an effective implementation strategy, Information and Management 29(4), 1995, pp. 215±225. [29] J.L. Whitten, L.D. Bentley, V.M. Barlow, System Analysis and Design Methods, 3rd edn., Burr Ridge, Illinois, Irwin, 1994. [30] R.E. Yellen, System analysts performance using CASE versus manual methods, Proceedings of the Annual Hawaii International Conference on System Sciences, 1990, pp. 497±501..

(26) 112. Prof. H.-L. Yang / Information & Management 35 (1999) 89±112. Heng-Li Yang received the B.S. degree from National Chiao-Tung University, Taiwan, in 1978, the M. Commerce degree from National Cheng-Chi University, Taiwan, in 1980, the M.S. degree in computer science from Pennsylvania State University, in 1985, and the Ph.D. degree in management information systems from the University of British Columbia, Canada, in 1992. Currently, he is a professor in the Department of Management Information Systems, National Cheng-Chi University. Formerly, he. was an associate professor at the National Cheng-Chi University and National Taiwan Institute ofTechnology. He also worked as a researcher in a ®nancial organization, a system analyst and a project manager in some information companies. His research interests include data & knowledge engineering, database and knowledgebased systems, software engineering, information management in organizations, technology impacts on organizations, and empirical studies in MIS. His articles have appeared in journals such as Information & Management, Journal Processing and Management, Cybernetics and Systems, Data and Knowledge Engineering, Expert Systems with Applications, Journal of Information Science and Engineering, and MIS Review..

(27)

參考文獻

相關文件

Root the MRCT b T at its centroid r. There are at most two subtrees which contain more than n/3 nodes. Let a and b be the lowest vertices with at least n/3 descendants. For such

• We will show a case study on constructing cylindrical panorama using a direct method.... Warp to

CASE (Center of Advanced Medical Simulation and Education) is a multidisciplinary medical simulation center aims to give high quality education with new training methodologies by

In Case 1, we first deflate the zero eigenvalues to infinity and then apply the JD method to the deflated system to locate a small group of positive eigenvalues (15-20

Students were required to compare in the formulation stage as the case teacher asked them to look at additional mathematical relationships, whilst they were required to compare in

HPM practice in Taiwan: A case study of HPM Tongxun (HPM Newsletter). These articles have documented the process of development and evolution of HPM practice in Taiwan as well

Then, it is easy to see that there are 9 problems for which the iterative numbers of the algorithm using ψ α,θ,p in the case of θ = 1 and p = 3 are less than the one of the

At least one can show that such operators  has real eigenvalues for W 0 .   Æ OK. we  did it... For the Virasoro