$MESO Started Phase 2/3 Covid ARDS trial 3 days ago with 6 locations now expanded to 14!, page-59

...Firstly, why are we using a Chi-square statistic to analyze group differences, when this is a non-parametric statistic that assumes the data distribution is not normal, as well as assumes violation of other parametric assumptions (eg linearity, HOV independence in the data etc... ?), without having actually tested for these?

We can still use parametric tests when the dependent variable is measured at a nominal level, and we are more likely to detect a significant effect when one truly exists, because of their statistical power.

Secondly, if we're going to use the Chi-square, the chi square value needs to be derived and interpreted, not just the alpha significance level. This should then be compared against the value for the set significance alpha level at the corresponding degrees of freedom (d.f) - we cannot just rely on the p values alone when looking at chi squares.

Using your 1st, 2nd and 3rd scenarios respectively, where p values are 0.05 and 0.001, for df=1, we would expect to see chi Sq values of less than or equal to 3.841 and less than or equal to 10.827, in order to verify statistically significant differences and accept/reject certain hypotheses. This would then need to be followed up by calculations of strength e.g Cramer's V, for significant Chi-square results obtained.

There is a bunch of other considerations that need to be factored in and calculated... And so it is a very time consuming process.