I think the estimates are as good as you will get with a probit model.  I think the overdispersion is due to the steep jump from the first dose level to the second, and then the relatively linear response after that, with little indication that an upper plateau has been reached with this dataset.  There are several examples of overdispersed logit/probit models being fit with PROC NLMIXED - perhaps that would be worth exploring.  If you get similar values for EC50 from both analyses, then you can have some reassurance that the PROC PROBIT results are acceptable.   SteveDenham ... View more

I should probably delete the earlier post, as it did NOT use the FREQ statement.  Here is code for a 2 parameter and a 3 parameter weibull.  The 2 fits the earlier points better, the 3 fits the later better (see attached output file).   proc reliability data=los; class treatment; distribution weibull; model length_of_stay = treatment ; probplot length_of_stay = treatment; freq freq; run; proc reliability data=los; class treatment; distribution weibull3; model length_of_stay = treatment ; probplot length_of_stay = treatment; freq freq; run; The median values for the 3 parameter are 7.3 for Yes and 7.5 for No, 10.1 for Yes and 10.0 for No for the 2 parameter, with substantial overlap of the 95% CI's.  I am still mystified as to why the median test in NPAR1WAY comes up significant.   SteveDenham     ... View more

All, please ignore this post, as it is incorrect.  Steve   Another way to approach this:   Suppose the length of stay is Weibull distributed.  You could use PROC RELIABILITY (if you have SAS/QC licensed) to get the fitted medians for each group, and look at the confidence bounds for those.  This code would do that (code edited, to what was used to generate the results below):   proc reliability data=los; class treatment; distribution weibull; model length_of_stay = treatment ; probplot length_of_stay = treatment; freq freq; /* this was in the code I used, as it is necessary to get the correct values */ run; The probability plots support the use of the Weibull distribution.     Results regarding the medians:       % Median Std Error 95% LOWER 95% UPPER Treatment 50 76.0081205 6.33476321 64.5532478 89.4956425 No 50 77.0518411 5.91355892 66.2910929   89.5593354   Yes     This tends to support the Wilcoxon results rather than the Median test results.   Edit: Thanks to @FreelanceReinh  for pointing out that the FREQ statement is needed, and wasn't in my code. SteveDenham   ... View more

Always take what's on Wikipedia with a grain of salt, but in this case, it makes a key point.  From the Median test page (bold emphasis added by me):     It is crucial to note, however, that the null hypothesis verified by the Wilcoxon–Mann–Whitney U (and so the Kruskal–Wallis test) is not about medians. The test is sensitive also to differences in scale parameters and symmetry. As a consequence, if the Wilcoxon–Mann–Whitney U test rejects the null hypothesis, one cannot say that the rejection was caused only by the shift in medians. It is easy to prove by simulations, where samples with equal medians, yet different scales and shapes, lead the Wilcoxon–Mann–Whitney U test to fail completely.   I am pretty sure that something like this is what is going on with these data - the shapes are different enough that the p values from the two tests don't align well.  It is even more interesting (at least to me) that the Median test comes out as "significant" while the Wilcoxon does not.   SteveDenham ... View more

@PaigeMiller shows that you can get what you want using PROC SURVEYMEAN, but you can also get the CLM with a DATA step by exponentiating the confidence bounds of the log transformed data.  STD and SE are a different matter.  Read through Alex Kritchevsky's webpage.  It has formulas for calculating the variability estimates, and lots of cautionary items about the use of these measures.  See http://alexkritchevsky.com/2018/06/15/geometric-mean.html    Alternatively, you could use PROC FMM to fit a non-mixture distribution of a lognormal distribution.  Use the OUTPUT statement to get a dataset with means and variances, and from those calculate standard deviations and standard errors.   In the end, though, @PaigeMiller  has the answer you requested.   SteveDenham ... View more

As no subject is specified for the R side variance component, the result is a measure of overdispersion..  I think this page of the GLIMMIX documentation covers your question: https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.5&docsetId=statug&docsetTarget=statug_glimmix_details05.htm&locale=en  In this case, the approach lifts the constraint on the scale parameter for the binomial distribution, which is ordinarily fixed at 1.  If your result is close to 1, I would think there is little evidence for overdispersion.  You could check the information criteria and the chi square/df ratio to see what is going on.   SteveDenham ... View more

From the QLIM documentation (the first example - Ordered Data Modeling), I found this:   By default, ordinal probit/logit models are estimated assuming that the first threshold or limit parameter (mu1) is 0. However, this parameter can also be estimated when the LIMIT1=VARYING option is specified.     There is code immediately following that looks like   /*-- Ordered Probit --*/ proc qlim data=docvisit; model dvisits = sex age agesq income levyplus freepoor freerepa illness actdays hscore chcond1 chcond2 / discrete(d=normal) limit1=varying; run; My first thought about this would be that your IV's are not all interval variables, and so there are going to be levels of the categorical variables set to 0.   You might consider using the OUTPUT statement to get marginal values or predicted values, and then post-processing those.   SteveDenham   ... View more

This note could be extremely useful: https://support.sas.com/kb/60/240.html    It shows what to do with the zero values obtained for parameters under LASSO.   SteveDenham ... View more

A lot is going to depend on the data you are fitting.  Can you share that dataset?   SteveDenham ... View more