Study Variables

Demographic variables include variables of age, gender, years of education, marital status (single versus married), occupation (with versus without occupation), and age of onset of psychotic symptom. Note that the married marital status consists of people living together and people getting married; housewives, students, people who never worked, who are unemployed or who already retired are included in people without occupation.

Environmental Factors

In this study, the environmental factors are related to obstetric complications, prena-tal growth retardation, special personal behavior, the psychological problem, and so on.

There are three environmental factors, described as follows separately.

(1) The patient has brain injury in the growth, such as prenatal growth retardation, brain damage, retarded intelligence and so on.

(2) Before getting disease, the patient had the unstable mood or abnormal behavior to interfere with adapting to the daily life, including angry, timid, depressed, inactive, having behavior problems, and so on.

(3) Before getting disease, the patient had the psychological problems to interfere with adapting to life in their infancy, including bad relation between parents, getting along badly with sibling or parents, getting disease about body, unforeseen happen-ings of family, and so on.

The first environmental factor was rated by a 3-point scale with 0 as no circumstance, 1 as slight (have not obviously heart body obstacle) and 2 as obvious (have obviously heart body obstacle). Due to the ratio of obvious subjects with the first environmental factor was too low, we combined the slight subjects with the obvious subjects in the first environmental factor. The others were rated by a 3-point scale with 0 as no circumstance, 1 as slight (have not obviously influenced routine life) and 2 as obvious (have obviously influenced routine life). There were one dummy variable for the first environmental factor, two dummy variables for the others.

Neuropsychological Variables

The neuropsychological battery assessed reaction time, attention, speed of informa-tion processing, and active problem solving. Specifically, the test battery included several standard neuropsychological instruments with demonstrated reliability and validity, in-cluding CPT, WCST, WAIS-R, WMS-R and Trail Making Tests A and B. These tests are briefly described below.

• Continuous Performance Task (CPT; Rosvold et al., 1956).

We used a CPT machine from Sunrise Systems, version 2.20 (Pembroke, MA, USA).

The procedure has been described in detail elsewhere (Liu et al., 1997; Chen et al., 1998a).

Briefly, numbers from zero to nine were randomly presented for 50ms each, at a rate of one per second. Each subject undertook two CPT sessions: the undegraded 1-9 task and the degraded 1-9 task. During the undegraded session, subjects responded to the target stimulus (the number 9 preceded by the number 1) by pressing a button. A total of 331 trials, 31 of them targets, were presented over 5 min for each session. During the degraded session a pattern of snow was used to toggle background and foreground dots so that the image was not distinct. The sensitivity index (d⁰) of the CPT performance reflects the subject’s sustained attention. Hence the CPT d⁰ was employed in this study as an external validation indicator of the subjects.

• Wisconsin Card Sorting Test (WCST; Heaton et al., 1993)

The Wisconsin Card Sorting Test is a commonly administered neuropsychological test sensitive to frontal lobe impairment, difficulties in information processing, concept forma-tion, and flexibility of abstract thought. For the purposes of this study the perseverative error score and the number of categories completed were used.

• Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler, 1982)

The WAIS-R is a standardized measurement of adult general intelligence. For this study was used the Full Scale IQ to explain the correlation between the structures and intelligence.

• Wechsler Memory Scales-Abbreviated (WMS-R; Wechsler, 1987)

The overall WMS-R battery is a comprehensive set of tasks designed to quantify encoding and retrieval processes. This study used a Total score which is the sum of WMS-R Logical Memory I and Logical Memory II.

• Trail Making Test (TMT)

The TMT provides information on visual search, scanning, speed of processing, men-tal flexibility, and executive functions. Originally, it was part of the Army Individual Test Battery (1994) and subsequently was incorporated into the Halstead-Reitan Battery

(Reitan & Wolfson, 1985). It consists of two parts. TMT-A measures the speed at which a subject to draw lines sequentially connecting 25 encircled numbers distributed on the sheet of paper. TMT-B measures the speed at which a subject can connect 13 numbers and letters in alternating sequence (1, A, 2, B, 3, C, etc.). The time needed to complete each task is recorded.

3.2.4 Regression Extension of Latent Class Analysis (RLCA)

Latent Class Analysis (LCA) is a statistical method for finding subtypes of related cases (latent classes) from multivariate categorical data. It can be used to find distinct diagnos-tic categories given presence/absence of several symptoms, types of attitude structures from survey responses, consumer segments from demographic and preference variables, or examinee subpopulations from their answers to test items. As with other latent vari-able models, like factor analysis, LCA is a procedure that attempts to explain covariation among a set of observed variables, by modeling the covariation of observed variables with unobserved (and hence latent) variables, that are fewer in number than observed ones.

The results of LCA can also be used to classify cases to their most likely latent class.

RLCA (Huang and Bandeen-Roche, 2004) extended the latent class model to allow both the distribution of the underlying class variable and the within-class distributions of mea-sured indicators to be functionally related to individual-level independent variables. It is assumed that the observed indicators are related to each other only through the latent variables. For example, within a latent class that corresponds to a distinct medical syn-drome, the presence/absence of one symptom is viewed as unrelated to presence/absence of all others.

Unlike factor analysis, RLCA is designed for use with dichotomous (or polychotomous) variables and assumes that the latent variables are also categorical. RLCA is used in way analogous to cluster analysis. That is, given a sample of cases (subjects, objects, respondents, patients, etc.) measured on several variables, one wishes to know if there

is a small number of basic groups into which cases fall. Briefly, RLCA works as follows:

The data required for input consist of the frequencies of all possible cross-classifications of the observed. RLCA then uses maximum likelihood estimation to fit one or a series of hypothesized models to explain covariance patterns among the observed indicators.

The parameters of RLCA are: (1) the prevalence of each of J latent classes, which are η_j(x_i) where x_i is a P × 1 vector of covariate and j = 1, · · · , J ; i = 1, · · · , N , and (2) conditional probabilities for each combination of latent class, item or variable (the items or variables are termed the manifest variables), and response level for the item or variable, which are p_mkj where m (= 1, · · · , M ) is the mth items or variables and k (= 1, · · · , K_m) is the kth level of the mth items or variables, that a randomly selected member of that class will make that response to that item/variable. The latent class probabilities provide information about the frequency of occurrence of each latent class. The latent conditional probabilities provide information about the degree of association between each of the observed variables and the latent classes, and are analogous to factor loadings in factor analysis (McCutcheon, 1987). Conditional probabilities give the sensitivity of the observed variables for indicating a particular latent class. Further technical details about parameter estimation and other aspects of RLCA can be found in Huang and Bandeen-Roche (2004).