Analysis method Strengths Limitations Notes
The van Elteren test and the Tadap2 tests No underlying assumptions about the distribution of the observed data are necessary.
It is possible to adjust for stratum effects.
Only possible to adjust the data for one stratification variable. Only two intervention groups can be compared. The Tadap2 test is available only in a code written in R. Non-parametric test. Often the optimal choice of analysis method.
Bootstrapping No underlying assumptions about the distribution of the observed data are necessary. The uncertainty of the non-parametric confidence interval provided via bootstrapping could be large if it is based on a limited number of observations. Non-parametric method to estimate the confidence interval of, e.g., means difference or median difference of two treatments.
The Kruskal–Wallis tests Multiple intervention groups can be compared.
No underlying assumptions about the distribution of the observed data are necessary.
The stratified version of the test is available in R only. Non-parametric method.
A valid method but only if more than two groups have to be compared.
The Wilcoxon rank sum test No underlying assumptions about the distribution of the observed data are necessar Not possible to adjust for stratum effects. Non-parametric test. The van Elteren test is often a better choice.
Generalised linear model Trial results can be adjusted for multiple stratification variables.
Odds ratios and confidence intervals can demonstrate the intervention effects.
Multiple intervention groups can be compared.
It can be impossible to deal with some data due to overdispersion[12]. Model checking will often show that the underlying assumptions of the models are not fulfilled. Parametric method. Often not an optimal choice of analysis method.
Generalised linear mixed model Trial results can be adjusted for multiple stratification variables.
Odds ratios and confidence intervals can demonstrate the intervention effects.
Multiple groups can be compared.
Can often handle overdispersion.
Complicated analysis. Checking the underlying assumptions is not easy. Parametric method. Often not an optimal choice of analysis method.
Table 2: An overview of the strengths and limitations when using the different count data analysis methods.