Significance testing bysorting weeks with two groups

[Sander Soet]

Dear all,
I have a problem with the significance testing of my results. I want to apply non parametric tests to my results. I have basically two groups to compare: an equal weighted returns group and a value weighted returns group. I calculated the value weighted returns bysorting for weeks and using the _gwtmean package. However, the thing is that i want to write down the results and statistically check them per week too. Nonetheless, the value weighted returns are calculated per week and so are averages of the equal weighted returns. So for example, week 2 has for every of 156 companies the same value weighted return, while the equal returns ofcourse differ per company.
Here is an example of my datasample for one company (id=1), this results last until week 156(3years).

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte Companyid int week double bharVCCO0711 float(bharsp600 wmean_bharvcco)
1  0                  1         1         1
1  1   .998989898989899 1.0293872 1.0242984
1  2 1.0101010101010102  1.030024  1.022209
1  3 1.0444444444444443  1.028824 1.0130448
1  4 1.0313131313131314 1.0333545 1.0274922
1  5 1.0676767676767676 1.0192732  1.033926
1  6 1.1383838383838383 1.0334525 1.0452584
1  7 1.2474747474747474 1.0352403 1.0391579
1  8 1.2121212121212122 1.0503502 1.0296736
1  9 1.2636363636363637 1.0714846 1.0251802
1 10 1.2363636363636363  1.033477 1.0239959
1 11 1.2717171717171716 1.0475339 1.0328152
1 12 1.2616161616161616  1.068154 1.0344493
1 13  1.197979797979798 1.0169221  1.038757
1 14 1.2676767676767677 1.0524563 1.0413563
1 15 1.2171717171717171  1.087966 1.0380343
1 16 1.1828282828282828 1.1079493 1.0513456
1 17 1.1727272727272726 1.0667336 1.0283052
1 18 1.1444444444444444 1.0880394 1.0190839
1 19                1.2  1.116594  .9990408
1 20 1.1525252525252525 1.0872067 1.0018867
1 21 1.1939393939393939 1.0997208   .990129
1 22  1.305050505050505 1.0918843  .9808657
1 23  1.314141414141414 1.0744967  .9703594
1 24 1.3202020202020202 1.0772395  .9714105
1 25  1.316161616161616 1.0353872   .980527
1 26  1.222222222222222 1.0276976  .9721831
1 27 1.3272727272727274 1.0529461  .9749035
1 28 1.3484848484848484 1.0767252  .9773502
1 29 1.3202020202020202 1.1135573    .98844
end

For the statistical check of the equal bhar(=returns) i used the following code: bys week: signrank bharVCCO0711=bharsp600 (=> sp600 is the benchmark)
This code is not possible for the value weighted bharsp600 and valueweighted bharvcco , because they were both calculated as means per week. I thought as a solution to perform this code:
bys week: signrank bharVCCO0711=wmean_bharvcco , but i am not sure if this tests counts the negatives and positives of the equal weighted versus the mean of the value weighted or differently. If the prior is the case i can check whether the equal and value weighted groups significantly differ per week and so still make conclusions about the significance.

Thanks a lot in advance!!
greetings Sander

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex (which you do). Note also that most of us are not from your area - you'll increase your chances of a useful answer by explaining in general terms.

I guess there are two ways to address this. First, you can ignore the fact that you have the same values for multiple firms within weeks, and just do the test. With non-parameterics, the order will still hold although it may overstate the sample. Alternatively, you can collapse the data and work at the week level. I don't know which makes sense - see what is done in your field.