Estimating asymmetric percentile from GLM

[SWEETSAS]

To estimate percentile for n observations, one can use say PROC MEANS, PROC UNIVARIATE etc. 

How can I estimate asymmetric percentiles  for an estimate from proc GLM? If the percentiles were symmetric one can use alpha=0.05 for instance to get 2.5th and 97.5 percentile of the estimates. But how can one obtain 20th and 90th percentile of an estimate from proc GLM? 

[PaigeMiller]

You'd have to go to a textbook which contains the formula for the specific confidence interval of interest, and program that yourself so you can then specify any percentile on the lower end and any percentile on the upper end.

Alternatively, you can run PROC GLM twice, for example once with alpha=0.40 and once with alpha=0.20