About kivester

kivester · ‎08-13-2025

I apologize for jumping into a thread, but is there a way to get a subscript capital L?

kivester · ‎04-21-2023

Thank you for your quick response, SteveDenham. That is very helpful!

kivester · ‎04-17-2023

I am looking for input on the most responsible way to report results. I regularly run models in Proc Glimmix to look at factors that effect exposure to particulate matter. Nearly without exception, running these models with the dist=logn is the most successful approach in terms of convergence and normal distribution of residuals. I struggle with the best way to represent the results. My audience and I are most interested in understanding the results in terms of the original scale: How does exposure differ between groups? What exposure might be expected for each group? I would ideally like to display mean and 95% confidence intervals. Is the best approach to achieve back transformed lsmeans (and back transformed standard errors in order to calculate 95% confidence intervals) as described by SteveDenham on these community boards? Is it okay to calculate the upper and lower confidence limits as I have coded? Or is the best way to convey findings to perform exp(estimates) and actually display the geometric mean, especially if I am including the confidence intervals in any figures and tables? My audience is usually not well versed with the intricacies of generalized linear models. Sample code below: Proc Glimmix data=extended plots=all; class Hay Measurement; model maxPM2_5=hay/ dist=logn solution; random horse; lsmeans hay/ cl plot=meanplot (cl); ods output lsmeans=lsmeans; run; data btlsmeans; set lsmeans; omega=exp(stderr*stderr); btlsmean=exp(estimate)*sqrt(omega); btvar=exp(2*estimate)*omega*(omega-1); btsem=sqrt(btvar); up=btlsmean+(1.96*btsem); low=btlsmean-(1.96*btsem); run; Thank you for any insight that you can provide.

kivester · ‎01-30-2023

Did you ever find the solution to this problem? I am in exactly the same boat, so far none of the proposed fixes are working for me!

kivester · ‎09-14-2022

Thank you for your help! You are correct, I do not expect that any timepoint will have values from 0 to 100. I will be happy to avoid the ordinal multinomial approach. Thank you for the Stroup reference! Katy

kivester · ‎09-02-2022

I need to perform a sample size calculation for a proposed study that the outcome of interest is a visual analog score on a scale 0 to 100. The score will be recorded at the start of the study (week 0) and weekly for 6 weeks. There will be three treatment groups. A time*treatment interaction is of interest, but we specifically would like to know the estimated sample size necessary to detect a difference between the 3 groups at week 6. To add to the complexity, the study will be conducted at 6 different centers. I have fairly good pilot data and expect the scores to look like this: Week0 Week1 Week2 Week3 Week4 Week5 Week6 Treatment1 42.2222222 65 75 78.75 85 85.625 87.5 Treatment2 31.4285714 46.4285714 51.4285714 61.4285714 67 67.1428571 70.7142857 Treatment3 36.8253968 50.9 55.9 57.775 60.9 61.2125 62.15 I know I can estimate sample size in proc GLMPOWER, but with the outcome on an ordinal scale I am not sure that is the best thing to do. Thanks for any help!

kivester · ‎10-16-2020

Thanks for your reply Steve. Regarding the random statement, how would you include week in the random statement and still specify r-side effects? As far as the "mixed" portion of my model, all I want to do is account for the repeated measures aspect of the design, and not spend much time looking at the statistical significance of this random effect. This seems to be a speed bump for some reviewers and in future I think I will describe my model as a repeated-measure generalized linear model. I am still a bit stuck on how to respond in this case as I don't believe that I have a serious problem with the model I have specified. Thanks for your help!

kivester · ‎10-15-2020

I am modeling the proportion of white blood cells (TNCC) that are neutrophils (PMN) recovered in bronchoalveolar lavage fluid as a function of diet (forage, 2 levels) at 4 time points using the following code: proc glimmix data=BAL plots=all; class number forage week; model freq_PMN/TNCC = forage week week*forage/solution link=logit; random_residual_/subject=number type=ar(1); run; This model has worked well for us in numerous experiments and does so here as well. In this instance the sample size is small with only 7 individuals as this was run as a pilot study. I am being queried about checking the assumption that random effects are normally distributed. I usually state that residual plots are visually examined to check appropriate designation of distribution/link function. (In this case they look good) But I guess I need something to specifically address this reviewer's concern about the distribution of the random effect (repeated measure). I have searched this topic in general, and most publications state that this is difficult, and depending on the article, either has important implications or doesn't. I would appreciate any guidance!

kivester · ‎04-27-2018

I have a series of graphs that I have created (in both proc genmod and proc plm, sas v 9.4) that I have then modified with the ods graphics editor. I need to export these graphs as eps files. I thought that I should be able to accomplish this with sgrender, but cannot manage to get the EPS file format. Help would be GREATLY appreciated! Katy

kivester · ‎12-09-2017

Wonderful! Thank you!

kivester · ‎12-07-2017

@sld, Thanks for your help. I haven't read all of the links that you have provided yet, but to answer some of your questions, my actual dataset has 67 observations. Yes, I have some subjects that I have only one observation for, and others that I have 2 or 3 observations. The instances of PMN>TNCC is an accident of the random data I used. This does not occur in my data set. Yes, I am happy to fit a random intercept model. I have included the random _residual_ statement due to apparent over dispersion. Thanks! Katy

kivester · ‎12-07-2017

@Rick_SAS, Yes, that makes sense. Is there a way to specify that the predicted values in the output are for trainer 1 and the mean values of the covariates? I realize the fit won't be very impressive, but this type of graph has been requested by my co-authors. Thanks again for your help! Katy

kivester · ‎12-06-2017

Rick, Here is some made up data. Thanks for agreeing to help! Data MadeUP; Input horse Freq_PMN TNCC trainer resp age inhal lps glucan; datalines; 1 6 300 8 0.35 4.2 0.34 1.4 53.1 2 4 112 8 0.064 4.416 1.492 9.143 256.411 3 20 189 4 0.191 0.137 3.105 5.427 262.197 4 17 291 4 0.124 4.188 3.126 4.923 208.315 5 18 385 1 0.058 2.956 3.289 0.945 254.197 5 0 123 1 0.025 0.464 1.784 7.311 283.147 7 7 226 2 0.175 3.965 0.668 6.183 186.320 8 17 15 4 0.171 1.774 3.091 4.858 121.950 8 17 320 4 0.148 1.699 1.088 9.965 257.282 8 2 1 4 0.129 2.507 1.012 7.521 322.331 8 17 56 4 0.179 2.538 1.924 0.651 110.123 9 15 113 3 0.110 4.037 2.661 0.585 333.831 10 3 44 1 0.182 4.877 0.077 5.208 292.303 14 20 299 7 0.177 2.238 2.490 7.207 327.078 15 17 246 1 0.043 0.581 0.930 6.753 64.075 15 7 78 1 0.013 3.826 0.622 5.364 311.640 15 9 285 1 0.032 2.302 3.636 5.400 107.441 18 8 152 5 0.016 2.722 1.801 7.350 341.055 19 14 392 5 0.160 0.737 0.999 5.862 314.837 20 10 398 7 0.112 2.973 3.691 7.995 322.038 20 20 328 7 0.039 2.221 3.991 9.583 163.122 21 13 206 4 0.148 1.245 3.986 0.648 322.854 22 17 187 2 0.197 0.926 1.096 3.555 130.459 23 20 253 1 0.171 2.663 3.850 8.469 114.644 24 11 97 1 0.086 2.594 1.739 0.959 310.298 25 16 83 4 0.072 3.385 3.934 6.602 198.959 26 14 248 8 0.002 0.546 1.020 8.069 66.514 26 14 364 8 0.027 3.378 1.559 9.159 292.377 26 14 26 8 0.107 1.262 0.825 6.589 135.586 29 3 119 1 0.169 0.517 3.756 2.659 303.409 30 9 74 1 0.059 4.357 1.232 5.881 83.229 31 6 280 1 0.035 3.508 2.747 9.789 10.971 33 6 251 1 0.118 2.325 3.156 3.601 306.553 ; proc glimmix; class horse trainer; model freq_PMN/TNCC= trainer resp age inhal lps glucan resp*lps; random int/ subject=horse type=ar(1); random _residual_; run;

kivester · ‎12-06-2017

I am trying to do the same thing. I have a generalized linear mixed model with several covariates that I have modeled in glimmix: proc glimmix; proc glimmix; class horse trainer; model freq_PMN/TNCC= trainer resp age inhale ups glucan resp*lps ; random int/ subject=horse type=ar(1); random _residual_; run; I am interested in plotting the predicted and 95% CI for freq_PMN/TNCC versus resp at the mean levels of the other covariates and overlaying the observed values. Is this possible? Thanks!

kivester · ‎12-06-2017

Using sas v 9.4, I am trying to get my graphs in the correct format for a specific journal. All of the graph text including axis labels and tick mark labels need to be Arial 8 point font. I am able to create an editable graph so that I can open in ods graphics editor and easily change the axis label font and size, but not the size of the font for the tick marks with the following code: ods graphics on; ods listing style=statistical sge=on; proc glimmix; class trainer; model resp=trainer/ link=log solution; random int/ subject=horse type=ar(1); lsmeans trainer/ ilink cl adjust=tukey plot=meanplot(ilink); run; How can I change the font size for the tick marks? Thank you! Katy

Online Status	Offline
Date Last Visited	‎08-14-2025 06:34 PM

Re: How to get the subscript and superscript in proc sg plot.

Re: Reporting results from Proc Glimmix modeled under log normal distr...

Reporting results from Proc Glimmix modeled under log normal distribut...

Re: how to add a superscript on y axis label?

Re: Sample size for visual analog scale outcome in three-level treatme...

Sample size for visual analog scale outcome in three-level treatment t...

Re: Checking model assumptions in Proc GLIMMIX

Checking model assumptions in Proc GLIMMIX

How do I save an .sge file as .eps?

Re: Changing the size of tick mark labels proc glimmix mean plot.

Re: Reporting results from Proc Glimmix modeled under log normal distr...

Re: Changing the size of tick mark labels proc glimmix mean plot.

Re: How to plot regression line from mixed models in my observed value...

Re: How to get the subscript and superscript in proc sg plot.

Re: Reporting results from Proc Glimmix modeled under log normal distr...

Reporting results from Proc Glimmix modeled under log normal distribut...

Re: how to add a superscript on y axis label?

Re: Sample size for visual analog scale outcome in three-level treatme...

Sample size for visual analog scale outcome in three-level treatment t...

Re: Checking model assumptions in Proc GLIMMIX

Checking model assumptions in Proc GLIMMIX

How do I save an .sge file as .eps?

Re: Changing the size of tick mark labels proc glimmix mean plot.

Re: How to plot regression line from mixed models in my observed value...

Re: How to plot regression line from mixed models in my observed value...

Re: How to plot regression line from mixed models in my observed value...

Re: How to plot regression line from mixed models in my observed value...

Changing the size of tick mark labels proc glimmix mean plot.