p gain method

[u58497688]

Hello,

i am working with the p-gain method and wondering if someone has experience with the the method.

My data is build up like this and has the Variable name and p-values from single-variable correlations (A1 to An; B1 to Bn) and ratio-correlations (A1_B1 to An_Bn):

Variable   var1   var2    p_value   p_gain

A1_B1     A1       B1      0.02

A1_B2     A1       B2     0.3

...

A1            A1        .       0.06

A2            A2        .       1

B1             .         B1     0.0007

...

Where "variable" is a ratio between the two variables "var1" and "var2". For the p-gain method and therefore the new variable p_gain, I want to calculate the following:

- If the p-value from the first part of the ratio, so here var1 (f.e. A1= 0.06) is smaller than the p-value from the whole ratio (A1_B1 =0.02) then the p_gain-value should be calculated as p_value(A1) / p-value (A1_B1)

-  If the p-value from the second part of the ratio, so here var2 (f.e. B1= 0.0007) is smaller than the p-value from the whole ratio (A1_B1 =0.02) then the p_gain-value should be calculated as p_value(B1) / p-value (A1_B1)

Thanks!

[whymath]

It's about table lookup skill and if statement.

data test;
  input Variable$ var1$ var2$ p_value p_gain;
  cards;
A1_B1 A1 B1 0.02 .
A1_B2 A1 B2 0.3 .
A1 A1 . 0.06 .
A2 A2 . 1 .
B1 . B1 0.0007 .
;
run;

data want;
  set test;

  if find(Variable,'_') then do;
    do i = 1 to rec;
      set test(keep=Variable p_value rename=(Variable=tmpVariable p_value=tmppvalue))nobs=rec point=i;
      if var1 = tmpVariable then var1pvalue = tmppvalue;
      if var2 = tmpVariable then var2pvalue = tmppvalue;
    end;
    if . < var1pvalue < p_value then p_gain = var1pvalue/p_value;
    if . < var2pvalue < p_value then p_gain = var2pvalue/p_value;
  end;
  drop tmp:;
run;