Univariate or bivariate analysis

univirate Though little bit confusing, terms ‘univariate and bivariate’ are used interchangeably with sole aim of analyzing relationship only between two variables in the absence of other confounding variables. Henceforth, we will use term univariate for subsequent discussion. Depending upon sample size of study, distribution of variable of interest, and its data dispersion, univariate analysis may be parametric or non-parametric.

• Commonly cited parametric tests in research are t-tests (both independent samples and paired samples), one-way ANOVA (analysis of variance), repeated measure ANOVA, and Pearson correlation coefficient.

• Their non-parametric counterparts are Mann-Whitney U test, Wilcoxon signed rank test, Kruskal Wallis test, Cochran’s Q and Spearman correlation coefficient.

• Computation of univariate relative risk ratios and odds ratios is also a rite of passage when analyzing association between outcome and exposure variables in cohort and case-control studies respectively.

• Kaplan Meier survival analysis is also a non-parametric univariate technique of choice when analyze a time to event data between two comparison groups.