Well, I checked the MIL-STD for inspection and found the following:
For a lot of 310,000 and Normal Inspection the sample size should be 800 (315 for Reduced Inspection). Your 300 is slightly too small for even the less accurate Reduced Inspection.
In addition, the sample must be RANDOM. As you culled out some folks, you lost randomness.
It would also seem to me that, based on things I have read over the years, that there are LESS NT Workstation licenses than OS/2 licenses in use. If this is the case, then NT Workstation winning should be less likely than OS/2 winning.
For questions or comments, email the Forums Editor