I have a question regarding analyzing data from multiple (in my case,
I conducted two different experiments A and B, each with several
conditions. Among them, one condition was common between study A and B
(only slight differences in stimulus used). The result was also very
similar. In other words, I have 2 datasets sharing the common interest
and similar results. Number of subjects for each study is 15. A few
subjects are overlapping between the studies, but they are mostly
different. Since the results for the specific condition are
interestingly similar between two studies, I would like to publish
them. I wonder if it is statistically valid to combine the two
datasets and perform the conventional two-level GLM.
I have come across studies, in which data from multiple studies with
small number of subjects were pooled to increase statistical power.
This approach seems more practical for my data. However, I am not sure
if there must be an inclusion criterion for pooling data in this way,
e.g., only the datasets that lack enough power to show activation
should be pooled. If my datasets already have enough power to detect
activation independently, can I still pool them?
Also, a few of the previous studies have termed such analyses as
meta-analysis. I am not sure if that is the right word, because in a
typical meta-analysis, the results from several studies are pooled
before carrying out a random meta-analysis, accounting for
I would greatly appreciate your comments and suggestions regarding this post.