Without applying normalization, the 2 sample t test of random eye had the largest P value ( P = 0.09), and the LMM had a borderline significant P value of 0.065, whereas other parametric tests had significant P value ( P = 0.045 for GEE score test, P = 0.038 for GEE Wald test), although their mean difference in TBUT between SS group and non-SS group were similar.
← all excerpts
Evaluation of Statistical Methods for Clustered Eye Data with Skewed Distribution.
1
—
—