I have a sample composed of two subsamples (say public and private companies) and a certain relationship is insignificant in both subsamples. If this relationship is significant in the collective sample, how can this be interpreted? Could this because of the large sample size of the collective sample?
Yes, the larger sample size in the combined sample could be the source of such a result. In particular, in a regression where the effect in each of the smaller subsamples is "borderline" (i.e., approaching significance), the same effect size could be significant when evaluated with a larger N, due to the smaller standard error associated with that larger N.
Thanks Dr. Morgan. Is it acceptable to put such interpretation for the difference in results in the final report or should I increase sample size to get more easy-to-explain results?
- Badges
- Science method
- Similar topics
- Teaching