SAS Support Communities

jgreenberg321 · ‎08-05-2020

Hello, I am interested in plotting the predicted probabilities from an interaction I've created using proc genmod. I would like to have the x-axis be a categorical variable by year, so the probabilities are displayed as strait lines between years. I set year as a categorical variable in my model with 2011 (earliest year) as the reference. However, when I create the effect plot, the years are displaying out of order. Does anyone know if there is an easy way to set the ordering/levels of the categorical variable on the x-axis. For reference, example code is below. Thank you! proc plm source=genmod; effectplot interaction (x=year sliceby=group ) ; run;

jgreenberg321 · ‎07-28-2020

Ha! Thank you— and thank you for your help!

jgreenberg321 · ‎07-28-2020

Sorry- the qq plot looks pretty good. There is some deviation out at the upper end of the data but quite linear over most of the data.

jgreenberg321 · ‎07-28-2020

Well, it's a risk-adjusted rate per hospital per year. It's actually calculated as the total predicted number of events over the total number of expected, per year. Therefore the numbers are generally non-integer values.

jgreenberg321 · ‎07-27-2020

Thank you for the detailed explanation. To clarify the outcome, it's a risk-adjusted complication rate on the hospital level. I'm not directly measuring/modeling the number of 0's and 1's in this context. Regarding the rest of your explanation, that all makes sense. I have found that trying to add some complexity to these models does create some problems with convergence, as you mentioned. Practically, I am primarily interested in the fixed effect estimate of time, which is easier to model. Thanks again.

jgreenberg321 · ‎07-27-2020

Thanks. I have about 5-7 time-point observations per level of hospitalid. The overall distribution of complication rates is pretty normally distributed. I did try to model model as a binomial in glimmix just to compare, but realized that may be problematic since the outcome is not an integer value (i.e. it is not 0/1 but rather 2.4, for example). I noticed one other things that seemed unusual: if I am not misunderstanding, it seems that to use proc mixed and put a random coefficient on time, along with a repeated statement, the "time" variable needs to be entered in the class statement. That would seem unusual, because it would prevent testing a linear fixed effect on time. Am I mistaken? Thank you!

jgreenberg321 · ‎07-27-2020

Thank you both for the feedback. That helps to clarify a lot regarding the appropriate syntax structure. To follow-up a few points: 1) The trend does look approximately linear when plotted across time, and visually there does seem to heterogeneity across cluster. 2) Regarding the outcome variable: the variable is essentially a binomial proportion (i.e. proportion of cases that had a complication). I had initially planned to use Glimmix, but didn't when the overall plot of the complication rates looked very much normally distributed. Would you still recommend modeling the event as a binomial outcome? Is the structure of the syntax (i.e. the repeated statement) generally the same when using Glimmix rather than proc mixed? Thank you!

jgreenberg321 · ‎07-26-2020

Hello, I am trying to model change in average hospital complication rates over time using proc mixed. I've tried reading articles online regarding the appropriate syntax for specifying this but want to make sure I am interpreting the results correctly. proc mixed data=have; class hospitalid; model complicated_rate=time/solution chisq; random intercept time/type=un subject=hospitalid_index g vc; run; However, my questions are as follows: 1) I tried specifying autocorrelation with "type=ar(1)" in my random statement, and my AIC was substantially larger. I assume that is indicating the model is better fit without specifying autocorrelation? 2) How can I request a significance test on my random effects (i.e. does the effect of time vary significantly across hospitals)? Thank you in advance!

jgreenberg321 · ‎07-23-2020

Thank you for explaining. Just to make sure I understand, the type 3 fixed effects values are equivalent to the LR test (i.e. testing the model with vs. without the added variable)? Fortunately in my example, the LR difference between models test aligns with the AIC change between the more complex and simpler models. Unfortunately, I was not able to access that paper, though can relate to the sentiment of lacking good measures for assumption checking.

jgreenberg321 · ‎07-22-2020

Just to follow-up on this: do you happen to know how to conduct a likelihood ratio test using proc Glimmix? I see from the Glimmix documentation that is possible but I can't find any examples of exactly how to do that. That would be helpful for testing the need for a polynomial term. Is that something that would be done through the "General" term? Example code is below. Thank you! proc glimmix data=test method=laplace ; class hospitalid Race_Group insurance female proc_group (ref='1'); model major_comp (event='1') =age age_square confounder Race_Group insurance female proc_group/solution CL DIST=binary link=logit; random intercept/subject=hospitalid; run;

jgreenberg321 · ‎07-21-2020

Thanks. I'll try that, along with other transformations (e.g. square root). I suppose we could use something like an AIC value to judge if the model with the transformed variable is better than the untransformed? I imagine we could also just use sgplot to plot a loess cuve in the residual plot? Thanks again.

jgreenberg321 · ‎07-21-2020

Hello All, I am fitting a mixed effects logistic regression model using proc glimmix to model post-operative surgical complications. I am interested in running model diagnostics, particularly to evaluate for any significant non-linearity in continuous predictors (e.g. relationship between age and log-odds of the outcome). I have had difficulty generating relevant plots using the glimmix procedure. Does anyone have any experience with this, or know of any good reference articles that might be helpful? All of the usefulf articles I've found on mixed effect model diagnostics in SAS are for linear models. For reference, my code setup is below. Thank you! proc glimmix data=test method=laplace; class hospitalid Race_Group insurance female proc_type; model complication (event='1') = age comorbidity_index Race_Group insurance female proc_type/S CL DIST=binary link=logit ; output out=out pred(noblup ilink)=expected_prob pred(blup ilink)=predicted_prob; random intercept/subject=hospitalid_index ; covtest/wald; run;

jgreenberg321 · ‎07-13-2020

That makes sense. Thank you!

jgreenberg321 · ‎07-12-2020

Makes sense. Thanks!

jgreenberg321 · ‎07-12-2020

I see. So it returns a value of "1" for outcome when the whichn expression is not equal to 0. Very clever. Since I ultimately wanted to have outcome=1 if the "any_" statements are "1" or if another variable >=1, it looks like I can do the following: outcome=0 ne (whichn("1", of any_:) or var2 >=1);

Online Status	Offline
Date Last Visited	‎08-17-2020 08:02 PM

SAS Support Communities

Effectplot question

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Longitudinal growth model using proc mixed

Re: Diagnostics for mixed effects logistic regression

Re: Diagnostics for mixed effects logistic regression

Re: Longitudinal growth model using proc mixed

Re: Counting frequency by group and year

Effectplot question

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Re: Longitudinal growth model using proc mixed

Longitudinal growth model using proc mixed

Re: Diagnostics for mixed effects logistic regression

Re: Diagnostics for mixed effects logistic regression

Re: Diagnostics for mixed effects logistic regression

Diagnostics for mixed effects logistic regression

Re: Defining array without numbered variable names

Re: Defining array without numbered variable names

Re: Defining array without numbered variable names

Follow Us

What is...