Log and power

[Marcus Eklund]

Hi!

I have an analysis with a quite small sample (700) and if I include income as a continuous vaiable i have significant results but if I include log income than they show the same coefficients almost but not significant. I wonder if this is than only because outliers may inflate my variables when income is not logged, or if the logged version somehow reduces the power of the analysis?

Best,
Marcus

[Nick Cox]

It seems highly unlikely that you get the same coefficients, so I stop there. I suggest you give much more detail, such as a display of your results.
Also, the implication is that you have other predictors, so information on that would be helpful.

I guess you don't mean power in the statistical sense here.