I have a logistic model that will require a random effect, requiring GLIMMIX.  However, before adding the random effect, I wanted to make sure I was specifying my model correctly without a random effect in GLIMMIX by comparing it to the output from GENMOD and LOGISTIC.  My code is below.  All variables are dichotomous: random (ACASI vs. FTF), time (2 vs 1), discTrt (Yes vs No).  The three procedures all give the exact same parameter estimates and standard errors.  GENMOD and LOGISTIC use Wald Chi-square to give p-values and these are identical.  GLIMMIX uses t tests instead, but the p-values are extremely similar.  Great! There are differences in the Type 3 Analyses.  The differences between GENMOD and LOGISTIC are extremely small and are due to GENMOD using LR test and LOGISTIC using Wald test.  As expected with all dichotomous variables, in LOGISTIC, the parameter estimate p-values equal the Type 3 p-values, because they are both doing the same Wald test.  The parameter estimate p-values in GENMOD equal the parameter estimate p-values from LOGISTIC and the Type 3 LOGISTIC, again, because they are all using the same Wald test.  The only odd man out is the Type 3 GENMOD due to the use of LR.  I'm comfortable with this. However, GLIMMIX instead performs an F test.  There are several unexpected things with this F test.  Perhaps they are obvious and I'm just overthinking it.  First, with dichotomous variables, I would expect the Type 3 F test statistic to simply be the parameter t statistic squared, as the F stat has 1 num df and the same denom df as the t stat.  This is not the case for either main effect, but is the case (considering rounding) for the interaction.  Thus, the p-values for the parameter estimates and the Type 3 tests, which I expect to be the same for dichotomous predictors, are different.  If I remove the interaction from the model, then the F stats do equal the t stats squared. Second, as a consequence of the first problem, the Type 3 p-values for the two main effects are no longer similar to the Type 3 p-values from GENMOD or LOGISTIC.  Again, if I remove the interaction, then all three procedures agree. Something is going on with GLIMMIX when an interaction is included that is different from GENMOD and LOGISTIC.  I can conceptually understand Wald Chi-sq and LR test, so perhaps I'm just no understanding the F test GLIMMIX is using and how the interaction would affect it.  In my results, also below, this doesn't much matter, as the interaction is not significant and can be removed from the model.  However, I am worried that there could be a case where the interaction is significant and must be kept.  Could there be a circumstance where a main effect in LOGISTIC/GENMOD is significant but the main effect is not significant in GLIMMIX, or vice versa? This would change the interpretation quite a bit.  Any insight would be appreciated. ods select Tests3 ParameterEstimates; proc glimmix data = aq;   class random (ref = 'FTF') time (ref = '1') discTrt (ref = 'No');   model discTrt = random|time / dist = binary solution chisq; run; ods select Type3 ParameterEstimates; proc genmod data = aq descending;   class random (ref = 'FTF') time (ref = '1') discTrt (ref = 'No') / param = ref;   model discTrt = random|time / dist = binomial type3; run; ods select Type3 ParameterEstimates; proc logistic data = aq;   class random (ref = 'FTF') time (ref = '1') discTrt (ref = 'No') / param = ref;   model discTrt = random|time; run; GLIMMIX Parameter Estimates Effect random time Estimate Standard Error DF t Value Pr > |t| Intercept -1.1081 0.1949 741 -5.68 <.0001 random ACASI 0.4360 0.2636 741 1.65 0.0985 random FTF 0 . . . . time 2 -0.9257 0.2800 741 -3.31 0.0010 time 1 0 . . . . random*time ACASI 2 -0.4120 0.3918 741 -1.05 0.2933 random*time ACASI 1 0 . . . . random*time FTF 2 0 . . . . random*time FTF 1 0 . . . . Type III Tests of Fixed Effects Effect Num DF Den DF Chi-Square F Value Pr > chiSq Pr > F random 1 741 1.38 1.38 0.2403 0.2407 time 1 741 33.38 33.38 <.0001 <.0001 random*time 1 741 1.11 1.11 0.2930 0.2933 GENMOD Analysis of Maximum Likelihood Parameter Estimates Parameter DF Estimate Standard Error Wald 95% Confidence Limits Wald Chi-Square Pr > ChiSq Intercept 1 -1.1081 0.1949 -1.4902 -0.7260 32.31 <.0001 random ACASI 1 0.4360 0.2636 -0.0806 0.9526 2.74 0.0981 time 2 1 -0.9257 0.2800 -1.4744 -0.3769 10.93 0.0009 random*time ACASI 2 1 -0.4120 0.3918 -1.1798 0.3559 1.11 0.2930 Scale 0 1.0000 0.0000 1.0000 1.0000 Note: The scale parameter was held fixed. LR Statistics For Type 3 Analysis Source DF Chi-Square Pr > ChiSq random 1 2.76 0.0966 time 1 11.00 0.0009 random*time 1 1.11 0.2924 LOGISTIC Type 3 Analysis of Effects Effect DF Wald Chi-Square Pr > ChiSq random 1 2.7361 0.0981 time 1 10.9306 0.0009 random*time 1 1.1059 0.2930 Analysis of Maximum Likelihood Estimates Parameter DF Estimate Standard Error Wald Chi-Square Pr > ChiSq Intercept 1 -1.1081 0.1949 32.3077 <.0001 random ACASI 1 0.4360 0.2636 2.7361 0.0981 time 2 1 -0.9257 0.2800 10.9306 0.0009 random*time ACASI 2 1 -0.4120 0.3918 1.1059 0.2930 ... View more

I have a dataset that I am running with the following code.  There are two treatments and two time points.  Each animal has one of the treatments and both time points: before and after treatment.  Each time point has replicates, though not the same number at each time point. proc gLiMMix data = b initGLM initIter = 500 itDetails chol method = quad(qPoints = 10 qCheck qMax = 100 initPL = 500);   class animalID treatment time;   model soft = treatment|time / dist = lognormal solution cl covB(details);   random intercept time       / subject = animalID type = un solution cl g; run; The interesting thing is that if I reorder the random variables to be: random time intercept.. I do not get a valid objective function.  If I switch to the default method rather than quad, it runs both ways; however, the degrees of freedom on the time variable when ordered "intercept time" (7) are half those when ordered "time intercept" (14).  Maybe I just haven't thought about this long enough, but I don't understand why the syntactical order of the variables would matter.  Is this only an issue with the intercept?  Do I need to keep ordering in mind if I had other variables in the mix too? Thanks! Michael (Windows 7 Professional, SAS 9.4) Edit: This also happens with a log-transformed dependent variable and a normal distribution, as well as using proc mixed with the log-transformed dependent variable.  In addition, for "time intercept" there are 0 degrees of freedom for the overall fixed intercept.  It's almost as if when specified in this order, the random intercepts are using up all the degrees of freedom from the fixed intercept. Message was edited by: Michael Cooney ... View more

Hello, I see that some help has been given to those with g- and r-side specifications, so perhaps someone could help me.  I'm not having many problems with convergence, just in specifying a model that makes the most sense for this pilot study.  There are: 9 animals (animalID); 2 treatments (treatment😞 5 animals get placebo and 4 study drug; 2 tissue types (tissueType😞 Vaginal and Cervical; 2 times points (time😞 Before and after treatment, before = day 0, after = day 14. Each animal has a tissue sample biopsied from both types of tissue both before and after: 9 * 2 * 2 = 36 clusters (20 placebo, 16 study drug). Each biopsy is separated into 3 to 6 equal sized sub-biopsies, infected with a microbe, washed, cultured, and analyzed for the presence and amount of the viral protein. I have no data on which sub-biopsies were contiguous to which in the original biopsy. It is not the case that the same sub-biopsy is tested before and after treatment in a paired fashion; the two biopsies (and their sub-biopsies) are different tissue, and there can be a different number of sub-biopsies per cluster even from the same animal and tissue type (e.g. animal X, vaginal tissue, before cluster has 6 sub-biopsies, after cluster has 4). There is no missing data. E.g. treatment animalID tissueType time replicate Infected AUC Placebo P1 vagina before 1 0 0 2 1 2681.342 3 0 0 4 1 1695.816 after 1 0 0 2 0 0 3 1 14965.546 4 1 987.683 5 0 0 cervix before 1 0 0 ... ... ... 5 etc. etc. after 1 ... 3 P2 vagina before 1 ... 6 after 1 ... 5 cervix before 1 ... 4 after 1 ... 6 etc. etc. etc. etc. etc. etc. etc. Thus, we have two time points for each tissue type, but each time point is a cluster of values, not just one measurement.  We want to analyze the data both as binomial (presence or absence of protein) and log-normal (amount of protein).  The goal is to see if the treatment decreases the rate of infection (> 0 protein) and the level of protein and if the different tissues are affected different.  One immediate problem is that the log-normal part is actually zero-inflated.  I'll get to that later. To model the binary portion, I am struggling to figure out the appropriate RANDOM statements.  Because I have a different number of sub-biopsies per cluster, I don't think this can be modeled R-side at all.  If I only had one measurement per time point and only one tissue type, I'm pretty sure the model would be R-side: proc gLiMMix data = c order = internal plots = all;     class animalID treatment time;     model infected(event = '1') = treatment|time / dist = binary link = logit solution cl oddsRatio;     random time / subject = animalID(treatment) type = cs residual; run; If I add back in the fact that I have clusters of measurements instead of just one, I think it gets forced into the G-side and becomes:     random intercept / subject = time*animalID(treatment) type = cs; This seems to take care of the clustering by time and animal, but maybe ignores that the two times within the same animal are correlated? If I add back in the tissue type, do I just further interact my subject=? proc gLiMMix data = c order = internal plots = all;     class animalID treatment tissueType time;     model infected(event = '1') = treatment|tissueType|time / dist = binary link = logit solution cl oddsRatio;     random intercept / subject = time*tissueType*animalID(treatment) solution cl v type = cs; run; Does this ignore that the two tissue types within the same animal should be correlated? Now, the second question concerns the log-normal portion.  The researchers want to know whether the study drug reduces the amount of viral protein.  If the 0s are ignored, the data is log-normal.  Would it be appropriate to analyze only those samples with > 0 protein to see if the study drug lowers the amount of protein only for those who got infected at all?  Clearly this answers a slightly different question: if infected, are viral protein levels lower, as opposed to, are viral protein levels lower.  If I can just model the > 0 portion, then I imagine I would set up my random statement the same was as for the binary portion, just switching over to lognormal. proc gLiMMix data = c order = internal plots = all;     class animalID treatment tissueType time;     model AUC = treatment|tissueType|time / dist = lognormal solution cl;     random intercept / subject = time*tissueType*animalID(treatment) solution cl v type = cs; run; Thanks! ... View more

Hello! I was wondering if there is anyway to apply multiple comparison adjustments to the F and t test done by the CONTRAST and ESTIMATE statements in PROC GLM.  Options like bonferroni, tukey, scheffe, dunnett are available in the MEANS and LSMEANS statements, but not with the CONTRAST and ESTIMATE statements. Thanks, Michael ... View more

Hello, Is there any function or macro available that will return the mantissa and the exponent of a number if written in scientific notation? For example, if the number is 123456789 (= 1.23456789E+08), I would like to be able to return two numbers: 1.23456789 and 8 Thanks! ... View more