# compare 2 different sized multivariate samples

I have a two data tables like below. The dataset1 represents failed candidates. The dataset2 represents the successful candidates.

I want to know, by applying some inference statistics which var (variable) has significantly changed the outcome (success and fail).

I have performed descriptive statistics and calculated centrality measures for both data sets column wise. I have also applied Machine-learning approaches for getting feature importance (Decision trees). But is there any way to statistically get that.

Dataset1 (2310x500)

    var1        var2        var3    var4  var500
U1  29.64611415 1576    427.4145    227  532
U2  28.45677687 1902    389.0615    225  484.88
U3  25.58181295 1376    379.1692    200  462
U4  27.02453608 1652    373.9734    210  457.60
U5  24.92124297 1790    330.8919    191  494.8
U6  17.00408003 1272    268.7932    131  409.30
U7  20.37879655 1088    340.4145    171  322
U2310 29.49205  1606    387.9543    222  642


Dataset2 (310x500)

          var1     var2 var3    var4  var500
S1  116.6229945 6954    1910.8676   890 2408
S2  77.54127969 4630    1222.5122   596 1421
S3  85.92342243 5190    1417.5633   655 1818
S4  53.90084493 3472    854.3462    418 939
S5  8.761080115 608     149.72      66  242
S6  4.530369972 356     98.8049     33  124
S7  5.944583534 624     124.3685    45  104
S310    23.41567    1208    418.46   204    266

• As long as var1, 2 etc are the same in the two data sets, then you can merge them and use e.g. logistic regression. – Peter Flom Sep 14 at 14:55
• Yes the var1, var2 etc are same for both data sets. But each data set represents independent group. I will follow your instruction and implement logistic regression. Also would like to see what LDA has to say for my data. Thank you. – Biofreek Sep 16 at 1:49
• LDA could be useful but it makes some pretty stong assumptions – Peter Flom Sep 16 at 10:50