[Bruce Weaver]

Hi Natallia. First, I still do not understand why your mathematical colleague is advising use of McNemar's Chi-square when your primary interest appears to be in whether or not there is a treatment effect. (I assume the z-test for comparing the two McNemar Chi-squares is the square root of a Pearson Chi-square on a 2x2 table where the row variable is Treatment, and the column variable is direction of change, positive or negative.)

Second, I think this is now getting beyond the point were I (or anyone else) can be very helpful through the continued exchange of posts in a discussion forum. I think you to find someone local with appropriate expertise. I see there is a Unit of Epidemiology and Biostatistics at one of the universities in your city. Perhaps someone there would be able to advise you.

HTH.

[Natallia Lapitskaya]

That what I get:
tabulate ib1.r_treat ib2.PersonWithn_PPI1 if !missing(PersonWithn_PPI3)
factor-variable and time-series operators not allowed
r(101);
[CODE]
* Example generated by -dataex-. To install: ssc install dataex
clear
input int id byte r_treat float(PersonWithn_PPI1 PersonWithn_PPI3)
9 1 1 1
11 2 1 1
12 1 1 1
17 1 0 0
20 2 1 1
39 2 0 0
53 2 0 0
54 2 0 0
55 1 0 0
56 2 0 0
57 1 0 0
59 1 0 0
end

May be I making mess in preprocessing/data management? I started with a "long dataset" where every medication is coded by a line, then defined if persons gets PPI will get 1 in the new variable PersonWithn_PPI1) Would it helps if I show you an example of the original dataset?

Sincerely, Natallia

[Natallia Lapitskaya]

Not sure If I posted the dataset correctly:-)

* Example generated by -dataex-. To install: ssc install dataex
clear
input int id byte r_treat float(PersonWithn_PPI1 PersonWithn_PPI3)
  9 1 1 1
 11 2 1 1
 12 1 1 1
 17 1 0 0
 20 2 1 1
 39 2 0 0
 53 2 0 0
 54 2 0 0
 55 1 0 0
 56 2 0 0
 57 1 0 0
 59 1 0 0
 65 2 1 1
 68 2 1 1
 69 2 1 0
 74 1 0 0
 79 1 1 0
 82 1 0 0
 93 2 0 0
 97 1 0 0
101 1 1 1
104 2 0 0
105 2 0 0
117 2 0 0
121 2 1 1
122 1 0 0
136 1 0 0
146 1 0 0
147 2 0 0
153 1 0 0
154 2 0 0
165 1 0 0
166 1 0 0
167 1 1 1
170 1 0 0
172 1 0 0
173 1 0 0
177 2 0 1
178 2 1 0
192 2 1 1
193 2 0 1
197 1 0 0
199 1 0 0
203 2 0 0
209 2 1 1
213 1 1 1
221 2 1 1
224 1 0 0
229 2 0 0
231 1 0 0
233 1 1 1
234 1 0 1
239 1 0 1
243 1 1 1
245 2 1 1
247 1 1 1
249 2 0 0
270 2 0 0
282 1 0 0
287 1 0 0
288 2 1 1
293 1 0 0
306 2 1 0
310 2 0 0
317 1 1 1
332 2 0 0
335 1 1 1
340 2 0 0
341 2 1 0
352 2 0 0
353 1 1 1
354 1 1 1
358 2 0 0
364 2 1 0
366 2 0 0
367 2 0 0
373 1 0 0
375 1 1 1
377 1 0 0
380 2 0 0
381 2 0 1
390 1 1 1
397 2 0 0
400 2 1 1
401 1 1 1
402 2 1 0
408 2 0 0
411 1 1 1
412 1 0 0
416 2 1 1
418 2 1 1
421 1 1 0
428 2 1 1
429 1 0 0
433 2 0 0
435 1 1 1
440 1 0 0
442 2 1 1
446 1 0 0
449 1 0 0
end

[Natallia Lapitskaya]

Thank you Bruce. Closing this post-chat. Thanks for your advices.
Kindly, Natallia