Xu, Man posted on Tuesday, June 25, 2013 - 3:47 am
Dear Dr. Muthen,
I also need more decimals in the results (lots of association tests - need very high threshold for statistical significance). The problem is that I have hundreds of models I need to run (through R or STATA) and so far I have not found an add-on thing in R or STATA that allows reading in the Mplus results output saved in such a way.
However, with a R package, I managed to read in the regular results printed in output files, including the z-values. So I want to calculate the p-values from R, using the Z values I got from Mplus. In this situation, do you think I can use the normal -distribution function in R? My sample size is around 1000.
For example, for a z-value -1.771 from Mplus I type this in R
> 2*pnorm(-abs(-1.771))  0.0765607
This gives more decimal places, which is what I need for the log 10 conversion. In Mplus the p value is 0.077. There is no way for me to be absolutely sure whether the two are identical, because the z-value itself doesn't have more accuracy after 3rd decimal places either.
Do you think the p value from Mplus is using the same distribution as in R this way?