Please tell me why I'm wrong :)
I have a dataset collected in a 2 by 4 repeated measures design, where 2 separate groups of people were tested 4 times on the same task. I have analysed my data with nparLD (https://www.jstatsoft.org/article/view/v050i12) because it is non normally distributed.
On top of these results, my research supervisor wants to see whether the ordered data has a recurrent pattern that would be different between the 2 groups. meaning, does one group always do better on the first try, for example. To do so, I ordered the continuous data such that for each subject, instead of having 4 values, I have numbers representing the ordering of the continuous data. For example, subject 14's continuous data is [ 56 42 44 79 ] and their ordinal data is [ 2 4 3 1 ] meaning that their data at Time 4 > Time 1 > Time 3 > Time 2.
I analysed this transformed data (exclusively made of a combination of 1 2 3 4 for each subject) with nparLD but I'm having doubts whether this is statistically ok, or if this should not be done, given that nparLD already takes into account the repeated measures... and technically it already gives me a Time RTE (relative treatment effect) for my continuous data.
QUESTION: can I really do that (re-analyse continuous data transformed into ordinal)?
I know that these two analyses ask different questions, but I want to make sure that the second question is not artificially creating results that are addressed by the first test (on continuous data).