@Demographer wrote: 1-Sorry I'm not sure what do you mean exactly by "You can set the parameter estimates of the LOGISTIC procedure as the starting values of the GLIMMIX procedure." Do you mean that if the intercepts of the model without covariate are for 0.5 0.2 0.3, I should use parms (0.5) (0.2) (0.3)? Yes, that is what I mean. You can also additionally set the starting values of other parameters (e.g., regression coefficients of the variables in the model other than the intercept) as the estimates obtained from the previous LOGISTIC procedure call. @Demographer wrote: 2-4 categories occ_reduced Frequency Percent Cumulative Frequency Cumulative Percent HIGH 467666 36.55 467666 36.55 LOW 107108 8.37 574774 44.92 MED 601957 47.05 1176731 91.97 UNEM 102782 8.03 1279513 100.00     In fixed effects logistic regression models, when one category of a categorical indepenent variable has too few observations or even zero observation, complete or quasi-complete separation can result, which means that the maximum likelihood estimator does not exist. When running SAS programs, this may be exemplified by failure of convergence. I could not remember if the same applies to the dependent variable as well and am also not that sure if such phenomena can be directly carried over to mixed effects model. But I saw that the second and fourth category of the dependent variable in your model both contain less than 10% of the entire observations in the dataset. Given that your log mentioned plenty of amount of removal of observations in model building process, it might be the case that too few observations whose dependent variable fall into the second and fourth category are eventually utilized. You can verify my conjecture on your own. So, if probable, you may try to combine the second and fourth category and build the model again. If that still does not work or combining categories is not permitted, look through the independent variables to see if variables with categories having little or zero data points exist. If that still does not work, look through the independent variables to see if any continuous ones exist. If so, you might try categorizing them (e.g., transform the continuous variable of age into elderly vs. not elderly). If that still does not work... then I am sorry that I have no idea how to tackle this problem in this case. ... View more

Another paper (Calculating power in analysis of variance. (apa.org)) found on the book Statistical Power Analysis for the Behavioral Sciences | Jacob Cohen | (taylorfrancis.com) also discusses the topic of power calculation in mixed effect models. Of closer revelance to your question is another paper cited in this book: Optimum Sample Size and Number of Levels in a One-Way Random-Effects Analysis of Variance: The Journal of Experimental Education: Vol 41, No 4 (tandfonline.com). A table of optimum sample size or number of levels for the random effects model is included in the latter paper. I hope these information helps. ... View more

Thank you for your reply and explanation! ... View more

If you add the FULLSTIMER option to the start of your program, your SAS log will report on the memory usage of each DATA or PROC step. You can then see how much you require versus how much you have: options fullstimer; ... View more

Maybe there are too many ouput dumped into sashtml.htm. Try to close all of destination to save your resource. Put the following line at the top of your code: ods _all_ close; .............. ... View more

All right. I have got it. Thank you for your patient reply! ... View more

Thank you again! That is another wonderfully useful piece of information! I am going to try out your method the next time I encounter problems. ... View more

Thank you Koen! These are terrific! ... View more

Hello. I am personally working on time-dependent variables in Cox models as well and have happened to read papers involving the functional form of time-dependent variables. It has been stated in Multiple imputation in Cox regression when there are time‐varying effects of covariates - Keogh - 2018 - Statistics in Medicine - Wiley Online Library that functional forms as flexible as restricted cubic splines can be used for time-dependent variables in Cox models. The functional form you mentioned is in nature piecewise linear, so I think it is also plausible to adopt. There is another issue I would like to mention: I think that in your case, starting from splines instead of adopting a piecewise linear functional form directly is a better choice. This can convince readers of your research results of the underlying rationale with respect to your selection of functional form. Even if you try splines and eventually decide to adopt the piecewise linear functional form you wished to employ in the first place, the results generated from splines could be of great help in the selection of the number and position of knot(s). By the way, the paper I mentioned concerns dealing with time-dependent variables in the presence of missing data, an issue hardly mentioned in the statistical literature. If missing values are present in your data, then it is also of use in helping you dealing with them. A last note before I end my reply: I am sorry that I do not know the exact SAS code to realize piecewise linear time-dependent variables, but the paper I cited contains the authors' links of R codes for utilizing restricted cubic spline as the functional form of time-dependent variable. Maybe you can reference their code. ... View more

I am sorry that I have not done much work on frailty terms in Cox models. I was, on the other hand, not that informed of the rationale underlying the claimed deprecation of the use of log-likelihood in the presence of frailty terms. I think attaching links of the information you cited (including the claimed invalidity of AIC in the presence of fraility terms) is of use for other users in this forum to look further into the details if they feel that are not fully informed of the rationales after reading your post. I agree with @Ksharp on the validity of the likelihood ratio test in Cox models in the general setting. But caution should be taken in your analytic project, as you have pointed out the possible invalidity of the use of likelihood for Cox models in the presence of frailty terms. Likelihood ratio tests are based on the comparison of log-likelihoods. If the use of likelihood is deprecated, then perhaps the use of likelihood ratio tests are deprecated as well. I used the word "perhaps" because I have not searched for evidence personally, so you are suggested to search the web and find for evidence yourselves. If both likelihoods and AICs are not valid in this case, then there is another graphical tool for the assessment of goodness-of-fit of Cox models. That involves the plotting the logarithm of the Cox–Snell residuals against their log-log transformed probabilities. See page 215 of A Handbook of Statistical Graphics Using SAS ODS for a detailed code of how to do this in SAS. ... View more

In addition to plotting, there is another subtlety I am interested in. That concerns the generation of spline effects themselves. Based on your code, restricted cubic splines are created after the imputation is done. I wonder if you included spline effects when you are imputing. If not, I am also concerned upon how this may have an impact upon your analytical results, as the majority of research paperes (e.g., Multiple imputation of missing data under missing at random: compatible imputation models are not sufficient to avoid bias if they are mis-specified - ScienceDirect) have shown that omission of nonlinear and interaction terms in the imputation stage and simply creat them on-the-spot in the analytical stage will cause biased-toward-zero results (i.e., the regression coefficients are biased toward zero). But, there are also research papers showing that creating them in an ad hoc manner may be a better choice in certain circumstances (e.g., Navigating choices when applying multiple imputation in the presence of multi-level categorical interaction effects - ScienceDirect). I personally have not yet retrieved any research paper specifically showing how spline terms, as a special and rather sophisticated kind of nonlinear term, should be dealt with in missing data. After all, creating them on-the-spot is rather easy and certain researchers (me included) are prone to adopt it had it been proved to be a viable approach. ... View more

您好，今天我在刷数科赛仕微信公众号时看到有个帖子，说的正好就是您的问题的解决方案。文中给出的解决方案和@whymath给出的不同，是在SAS内部通过数据步（DATA step）进行的。这是链接，供您参考：SAS干货 | 同时输出宋体和新罗马字体。在此感谢创作此文的骆豪老师！ ... View more

@amyip wrote: From my reading, zero-inflated and zero-truncated data are not the same phenomena. In my case, all my subjects had at least one prescription, so our outcome of interest, days' supply of medication prescribed, has no zero values. I think if you slightly modify the concept of truncation proposed by Breen into y≥c instead of sticking to the original definition of y>c and choose zero as the constant c, then zero-truncation can be a source of zero-inflation. They are not identical entities, the former can be one of the source of the latter. This subtle difference might be more a philosophical than a statistical issue. On the other hand, if you wish to strictly apply Breen's definition, then zero-truncation is not a cause of zero-inflation, because only the y's that are strictly larger than zero are recorded in the data, which is your case. ... View more