Valid method for analysing if positive association between giving a drug and increasing ordinal scores I have a set of ordinal scale scores for a group of patients before and after an intervention (drug). I was wondering what would be a valid statistical method for analysing if there was a positive association between administration of the intervention and increasing ordinal scale scores. The number of scores before and after the administration of the intervention are never equal in my data set because it was not a controlled study. The ordinal scale used is a 62-item scale. Image is of example results from two patients to better describe the data - I have data for 14 patients in total.

 A: Given the comments, it sounds like you are at a very preliminary stage of research on the topic and an exploratory approach would therefore be sensible. Do not be put off by that fact that the data are not from an ideal study. See what the data can tell you.
You are looking for a possibly small signal in fairly noisy data that has variability in the number of observations per patient and per condition. That means that your inferences will have to be quite cautious, and the outcomes can, at best, point to areas for further investigation.
Try plotting the averages of before treatment and during treatment (I'd look at means and at medians). Join the dots within patient and see if the  differences are at all consistent.
You could, perhaps, treat the several measurements in each patient as technical replicates and therefore take an averages for each patient (before and during) and perform a paired t-test (or a permuations test) on the averages. (Treat the results with caution.) A more sophisticated analysis that weights averages with lots of values more heavily than averages with few values, but that is beyond my capability.
