Statistical Tests

Statistical Tests

1. PARAMETRIC

   • About
     • Level or scale of measurement is of equal interval or ratio scaling
     • The data are normally distributed OR can be transformed
       • If the variables are counts or derived they must be transformed
     • Uses actual observations or measurements
     • Often compare means and variances of interval scales
       • When comparing variances, each need to be homogenous
     • No loss of information
     • ROBUST if the sample size is LARGE
     • Greater statistical power than nonparametric
   • TESTS
     • t-test
       • One sample t-test
       • Independent sample t-test
       • Paired sample t-test
     • ANOVA
       • One way ANOVA
       • Multi-factorial ANOVA
         • Hierarchical (nested) design
         • Two-way crossed layout
           • With replication
           • Without replication

2. NONPARAMETRIC

   • About
     • Used on Nominal or Ordinal data in ADDITION to Interval and Ratio
     • Distribution FREE
     • Used on observations, measurements or values represented by assigned ranks
     • Compare medians of interval or ordinal scales
     • Counts or derived variables can be analysed directly
     • Loss of information
     • Small sample sizes can be used as normality is not a prerequisite
     • Have LESS statistical power than parametric equivalents
   • TESTS
     • 2 independent samples
       • Mann-Whitney U
       • Kolmogorov-Smirnov Z test
     • K Independent Samples
       • Kruskal Wallis Test
       • Median Test
     • 2 Related Samples
       • Wilcoxon Test
     • K related samples
       • Friedman Test