CHS 627: Multivariate Methods in Health Statistics

Review/Preview:
Predictors, Outcomes, Null Hypotheses, and Statistical Tests

 

Univariate (no or one predictor)

Multivariate (two or more predictors)

 

UNIVARIATE
Number and type of predictor variables	Type of outcome	Null Hypothesis (H0)	Test(s)
no predictor variable	normal		one-sample t-test
	non-normal		one-sample median/ sign test
	categorical		one-sample z-test for proportions binomial test
			Chi-square goodness-of-fit test
no predictors, no outcomes (i.e., looking for association or independence among 2 or more variables)	normal		Pearson correlation
			factor analysis
	one or more non-normal		Spearman rank correlation
	categorical		contingency coefficient
			odds ratio
one categorical predictor (2 categories)	normal		two independent sample t-test (if independent groups or random assignment to groups)
			paired t-test (if matched samples or repeated measures)
	non-normal		Mann-Whitney Wilcoxon rank sum test (if independent groups or random assignment to groups)
			Wilcoxon signed ranks test (if matched samples or repeated measures)
	categorical		Chi-square test (if independent groups or random assignment to groups)
			McNemar chi-square test (if matched samples or repeated measures)
one categorical predictor (3 or more categories)	normal		Oneway ANOVA (if independent groups or random assignment to groups)
			Repeated Measures ANOVA (if matched samples or repeated measures)
	non-normal		Kruskal-Wallis test (if independent samples or random assignment to groups)
	categorical		Chi-square test (if independent samples or random assignment to groups)
one continuous predictor	normal		simple linear regression
	non-normal		transformation
	categorical		logistic regression

 

MULTIVARIATE
One outcome, two or more predictors
Number and type of predictor variables	Type of outcome	Null Hypothesis (H0)	Test(s)
2 or more categorical predictors (2 or more categories each)	normal		Factorial ANOVA (if independent groups or random assignment to groups)
	non-normal		transformation
	categorical		log-linear analysis
2 or more continuous predictors	normal		multiple linear regression
	non-normal		transformation
	categorical		logistic regression
mixture of categorical and continuous predictors	normal		analysis of covariance (ANCOVA)
			general linear model
	non-normal		transformation
	categorical		logistic regression
Two or more outcomes, one or more predictors
Type of predictor variable(s)	Type of outcome	Test
categorical	normal	multivariate analysis of variance (MANOVA)
continuous		multivariate multiple regression
mixed categorical and continuous		multivariate general linear model

Page editor is Julia Hartman.

Last modified November 29, 2000.
The information in this site is current through August 15, 2000.

Although the authors of this Web site have made every reasonable effort to be factually accurate, no responsibility is assumed for editorial or clerical errors or error occasioned by honest mistake. All information contained on this Web site is subject to change by the appropriate officials of The University of Alabama without prior notice. Material on this Web site does not serve as a contract between The University of Alabama and any other party.