can a rank based inverse normal transformation of a continuous trait affect the significance/power

Question

[muralinmars]

Hi,

I have this question from the recent analysis that I have been doing.

I am meta-analysing a continous trait (say TRAIT1), that is inverse normal transformed. And one particular locus that was GW significant in an earlier effort (using log-transformed values) completely loses its significance in the current analysis.

The sample size is more or less the same.

Here is a paper that suggests Inverse normal transformation can reduce power in few instances.

Is this possible!

Thanks in advance. M