Do you really want to test that the accelerometers are the "same", or is it sufficient to test for "at least one accelerometer is clearly different from the others". The reason is that the first is an equivalence test where two one-sided tests (TOST), or some generalization of that, would be used (ANOM ?), and you have to state what the boundaries on the equivalence interval are. For the second, an ANOVA would suffice (although you may want to fit a generalized linear model, depending on the residuals and a histogram plot). The issue with the equivalence test approach is that there are 36 pairwise comparisons, all of which would have to pass the equivalence criteria in order to say that the instruments were the "same". If one of the accelerometers is accepted as a standard, the number of comparisons drops to 8, and that may be tractable.   SteveDenham ... View more

Rather than analyzing the calculated proportion, would it make sense to analyze the actual live counts? I ask that because it looks like you "seeded" the pots with equal numbers at the beginning.. If that is the case, consider a model like:   proc glimmix data=puncv.new plots=pearsonpanel (conditional marginal) method=laplace; class Month Location Source Cage; model live_count= Month Location Source /dist=poisson; random intercept /subject=Cage; run; From here, you could consider interactions between you main effects, or whether there are correlations over time.  And in the end, you can use the mean estimates on the original scale divided by the number introduced as a proportional survival.   SteveDenham   ... View more

btvar = back-transformed variance btlsmean = back-transformed expected value   I guess I am having a conceptual issue here. I jumped into modeling this because I had some prior information about the distribution of the variables in question, i.e., that they were log normally distributed. If you believe in the distribution, a distribution-free estimate of location is probably not needed. It just turns out that the geometric mean obtained from a sample from a lognormal distribution is a good approximation to the median.   So, are you going to make a defensible assumption about the distribution of your response variables?  if so, then GLIMMIX or GENMOD are appropriate. If you truly want to estimate the median, you should consider PROC QUANTREG. Work through the examples in the documentation..   SteveDenham ... View more

The Hahn and Meeker CI is based on the assumption that the data are normally distributed. If that were the case, there would be no need to consider using a non-parametric approach to BE.   SteveDenham ... View more

Some things to think about: Exponentiating the estimate obtained from GLIMMIX gives you the geometric mean, which is a good approximation of the median on the untransformed scale. I don't think there is a good way to get Q1 and Q3 estimates, but you might try creating a confidence interval on the estimate with an alpha of 0.25, and then exponentiating the endpoints. Better would be to simulate 1000 or so samples from the population defined by the parameters you have found, and look at the quartiles of that distribution.   SteveDenham ... View more

I like @PaigeMiller 's answer (LOGISTIC option in PROC POWER) but want to point out a particular pet peeve of mine. A post hoc calculation of power is pretty useless, except as a jumping off point for a planned future survey/experiment. Here is the reasoning that I was taught: You already know the p value of interest. Depending on your decision rule, then you either detected a difference (power=1) or you did not (power=0).  It's why effect size is a major consideration in your interpretation of results.   So, as a prospective tool, you can use the effect size observed, a desired power or desired sample size, and an alpha level, you can calculate the remaining free variable. But that is for a collection of new samples from a population with characteristics similar to those in hand, not for the current sample from such a population. As I mentioned above, you already know the power for your current sample.   And you have to consider that the error estimates and point estimates from your weighted survey analysis have a built-in finite adjustment, which PROC POWER doesn't have.  In my opinion, that means simulation is probably the best way to calculate prospective power/sample size.   SteveDenham   SteveDenham   ... View more

With a little jiggering, you can get the bioequivalence boundaries from the two one-sided Wilcoxon tests.  And interestingly enough, at least to me, these boundaries are dependent only on the number of observations involved, i.e. for any given sample size, there is one Wilcoxon value for the not greater than and one for the not less than tests. If your Wilcoxon value falls between these, then you have some indication of equivalence.   All this comes with a BIG caveat. If your raw data  has one or more "outlier" value(s), converting to ranks throws that valuable information away.  Since most equivalence trials are somewhat restricted in sample size, those extreme values are telling you that the distributions are likely not the same.   So, rather than rank (or any other) transformation, you might look into resampling methods to calculate the difference in means, the variance of that difference, and the resulting confidence interval.   SteveDenham ... View more

In most cases, a bootstrap SE will be less biased than a delta calculated SE, as the delta method essentially linearizes the Taylor expansion, so higher order terms that could account for skewness are ignored.  So, if you have a reasonable population to sample from, then it would seem that there would be less bias. However, the calculation of a mean value across the various subsamples makes the same kind of linearization. I think the key for SE's might be to work in the variance space and then calculate a pooled estimate of the SE by formula.  For anything resampling related, look through Rick Wicklin's blog:The DO Loop. Another good start would be  https://communities.sas.com/t5/Ask-the-Expert/How-Do-I-Use-the-Bootstrap-Method-in-SAS-Q-amp-A-Slides-and-On/ta-p/836226    SteveDenham ... View more

Blatantly stealing @PaigeMiller 's code, here is what I have:   data have; input id $ date_of_injection $10.; sasdate=input(date_of_injection,mmddyy10.); datalines; A 12/1/2015 A 1/19/2016 A 4/19/2016 A 5/17/2016 B 12/17/2019 B 2/3/2020 B 3/23/2020 B 6/15/2020 C 9/10/2015 C 10/16/2015 C 12/23/2015 ; run; data want; set have; by id; prev_date=lag(sasdate); if not first.id then do; interval=sasdate - prev_date; output; end; drop prev_date; run; Which gives output that looks like:   id date_of_injection sasdate interval A 1/19/2016 20472 49 A 4/19/2016 20563 91 A 5/17/2016 20591 28 B 2/3/2020 21948 48 B 3/23/2020 21997 49 B 6/15/2020 22081 84 C 10/16/2015 20377 36 C 12/23/2015 20445 68  Interval here is the number of days since the previous injection.  Week can be calculated as week=floor(interval/7), or whatever function you want to get the week.   SteveDenham     ... View more

I hope the results table you want is just an example of the format, as the Interval values do not match the days.  As an example, look at the last 2 values for ID = C.  16 OCT to 23 DEC = remaining days in October + all days in November + days prior to injection in December, which is 15 + 30 +22 =67, not the 42 in the table.    Other folks following this will be able to tell you how to convert data in date format to SAS date variables, which can be manipulated by standard arithmetic.   SteveDenham ... View more

When you say "will not run" that isn't very useful without seeing the log for that procedure. Please post the entire log starting with the call to PROC GLIMMIX down to the next run; statement.   I suspect that the issue has to do with declaring variables both fixed and random, but it could be anything.   SteveDenham ... View more

How would a Bayesian approach assist you (since that is what BGLIMM is for)? At this point, I do not believe BGLIMM (or PROC MCMC) supports the EFFECT statement to generate a multimember design.   SteveDenham  ... View more

In order to understand what you are attempting, I need to know why you are fitting five different spline terms to the same data and then placing all of them in the same analysis (spl spl1 spl2 spl3 spl4). Do the different splines reflect different hypotheses about the time by hba1c relationship? Is there any reason to believe that time as interpreted by the spline function has a fixed effect on the outcome, and you want to see if it differs by sex? As an example for that, look at this in the PROC GLIMMIX documentation: https://documentation.sas.com/doc/en/pgmsascdc/v_036/statug/statug_glimmix_examples09.htm Or is it possible that there is a single factor response in time, with the variance being what needs a spline fit.  As an example look at this in the PROC GLIMMIX documentation:   https://documentation.sas.com/doc/en/pgmsascdc/v_036/statug/statug_glimmix_examples20.htm    SteveDenham ... View more