Ordinal survey, 2 groups: looking for correlation between groups

#1
Hi I have an ordinal data set across 50 survey questions (same response scale for each question).

The data is non-normally distributed for each question, and isn't suitable for conversion to interval data.

Much simplified example attached.

I need to be able to test if there is a significant correlation between Group 1 and Group 2's survey answers, and I'm struggling to find the right test.

I have non-parametric, ordinal, non-normal data so have considered:
- matched pairs (I don't think the different groups answers count as 'matched pairs'?)
- Wilcoxon signed-rank (this seems to need numerical data across one variable?)
- Kendall's W (seems to only compare overall ranks? wheras I have separate questions)
- Friedman test (this seems the closest - e.g. http://www.statisticslectures.com/topics/friedman/ if I replace 'weeks' with my groups, but I need to somehow compare the means or medians of each group rather than each individual respondent?

The hypothesis suggests there will be a difference between the two groups but I am unsure of how to accurately test for that.

Any help would be much appreciated.
thanks in advance,
Ivan
 
#2
In case of interest, I ended up using Cohen's weighted Kappa method, and medians of each group's scores. This is perhaps not ideal but seems the best I can do as there seems to be no weighted kappa method where there number of respondents is > 2.