What is the most reliable procedure for examining variance for the following dataset, where different variables are being examined across four groups from the same survey (sample) n = 1850:

e.g. Group 1 (n =700): Score = 95

Group 2 (n = 450): Score = 48

Group 3 (n = 350): Score = 16

Group 4 (n = 380) : Score = 35

The above example approximates the actual data (attached) for the variable with the most responses (n = 195). The variables of potential interest range from n = 195 down to n = 20, but likely the cut off would need to be higher to be meaningful ( n = 50, n = 100? - this would account for 13 or 8 variables for further examination, instead of up to 20...).

In the initial analysis I have examined ten variables (cut off would then become n = 55). Some variables overlap in the discussion, so n = 8-12 seems reasonable overall. Still, I would like to report for up to 20 in the table.

When the above data is recalcluated into %:

Group 1 = 5.61; Group 2 = 5.72, Group 3 = 6.98, Group 4 = 4.90

I have been carrying out the preliminary analysis with the above relative frequencies (%), which is fine initially for highlighting differences, for example, above between Group 3 and 4. However, now I am looking to consolidate the data analysis with the correct procedure(s) reported to support the qualitative findings.

Thanks for clarification on the above example(s) and guidance on correct methods.