Alpha level and Power

[stronghold]

Advertisement - Members don't see this ad

Alpha level (not Beta) increases when power increases. But what about when the sample size increases? I know bigger sample means more power, however, I don't know if there is a relationship with alpha level

 

[notbobtrustme]

alpha is where you set your significance, eg p = 0.05. Increasing the sample size reduces the chance of a type 1 error (eg finding a difference when none exists), but it doesn't affect the alpha level; only the researcher can do that and better have a damn good reason why he thinks an alpha level of 0.1 or w/e is valid.

 

[mcloaf]

alpha is where you set your significance, eg p = 0.05. Increasing the sample size reduces the chance of a type 1 error (eg finding a difference when none exists), but it doesn't affect the alpha level; only the researcher can do that and better have a damn good reason why he thinks an alpha level of 0.1 or w/e is valid.

I can't imagine why they'd bump alpha up. I've only heard of making it smaller for gwas.

 

[notbobtrustme]

I can't imagine why they'd bump alpha up. I've only heard of making it smaller for gwas.

yea, def for GWAS. When you analyze that much data, p @ 0.05 is trivial. Some researchers bump up their alpha to 0.10 to make results publishable in less reputable journals. We had a researcher in our department when I was doing my PhD who was notorious for fudging his alpha values.