口腔病理科 On-Line KMU Student Bulletin
- 1 -
原文題目(出處)： Which is the correct statistical test to use?
Br J Oral & Maxillofac Surg 2008;46:38–41 原文作者姓名： Evie McCrum-Gardner
通訊作者學校： University Of Ulster, Shore Road, Newtownabbey BT37 0QB, United Kingdom
報告者姓名(組別)： 林煒倫(int B組)
報告日期： 97/12/19
內文：
z Introduction:
Choosing the appropriate statistical test:
1. To decide what scale of measurement your data is.
2. Consider the analysis required
z Three main type of scales of measurement
1. Nominal: categories but no order. Ex: sex, marital status, location of lesion.
2. Ordinal: ordered categories. Ex: pain, stage of tumor.
3. Interval: including age (years), weight (kg) or length of osteotomy (cm).
z Parametric assumptions
9 Examine the distribution of interval-scale data to check if they are normally distributedÆbell-shaped, symmetrical about the mean.
9 normality tests: Kolmogorov–Smirnov and Shapiro–Wilks.
z Comparison of two groups
A. two independent groups
(1) Independent samples t-test Æcompare sample means from two independent groups for an interval-scale
(2) Mann–Whitney U-test Æcompare two independent samples when data are either interval scale but assumptions for t-test (normality) are not satisfied, or ordinal (ranked) scale.
(3) Chi-square (χ2) testÆ compare proportions between two or more independent
口腔病理科 On-Line KMU Student Bulletin
- 2 -
groups or investigate if there is any association between two nominal-scale variables.
B. two paired groups
(1) Paired samples t-test Æcompare two sample means where there is a one-to-one correspondence (or pairing) between the samples.
(2) Wilcoxon signed rank test Æcompare two paired samples when data are either interval scale but assumptions or the paired t-test (normality of within-pair differences) are not satisfied or ordinal (ranked) scale.
(3) McNemar’s test Æcompare two paired samples when the data are nominal and dichotomous.
z Comparisons of more than two groups
A. independent groups
(1) One-way analysis of variance (ANOVA) Æmore than two independent groups being
Compared if the parametric assumptions are satisfied.
(2) Kruskal–Wallis one-way ANOVA Æused for ordinal data, or an interval-scale variable, which are not normally distributed.
(3) Chi-square (χ2) test Æ used for nominal data.
B. paired/related groups
(1) Repeated-measures analysis of variance Æused if the parametric assumptions are satisfied.
(2) Friedman’s test Æused for ordinal data or an interval-scale variable that is not normally distributed.
(3) Cochran’s Q-test Æused for nominal dichotomous data when there re more than two related groups.
z Association between two interval or ordinal variables
9 correlation coefficient is used to investigate the association between two interval or ordinal variables.
1. Pearson’s product-moment correlation coefficient is used for both variables are interval and approximately normally distributed.
2. Spearman’s rank correlation is used for either variable is ordinal or interval and skewedcoefficient.
9 multiple regression analysis with an interval-scale response (dependent) variable and several
predictor (independent) variables.
9 Logistic regression is used when the response variable is dichotomous.
z Power, P values, and percentages
9 Make a sample size calculation so that the study will have sufficient power to detect significant difference.
口腔病理科 On-Line KMU Student Bulletin
- 3 -
9 It is often necessary to combine categories so that there are sufficient numbers in each group.
9 It is essential to also quote the numerator and/or the denominator so that it is clear how the percentage has been calculated.
9 A significance level (P value) is considered significant if it is less than 0.05.
題號 題目
1 比較不同種族之間得到鐮刀型貧血的統計方法中,樣本的分類是屬於:
(A)Nominal (B)Ordinal (C)Interval
(D)none of above
答案(A) 出處：
題號 題目
2 在統計學上, P值須在何種範圍內才可被認為是顯著的結果
(A)<0.05 (B)<0.5
(C)0.5~1 (D)>1
答案(A) 出處：
