Re: Mixture model with logistic regression

________________________________ From: Jeroen Elassaiss-Schaap (PD-value B.V.) <[email protected]> Sent: Saturday, February 20, 2016 4:35 PM To: Mark Sale Cc: [email protected] Subject: Re: [NMusers] Mixture model with logistic regression Hi Mark, Is it indeed a logistical model or is it an ordered categorical? I assume you refer to the latter. Not sure how you get your second category otherwise. Anyway, to me it reads like you are trying to have the mixture model describe exactly what the omega is trying to describe. Perhaps you could drop the omega all together? (or fix to a small value) I also like Bob's suggestion, I would go for it ($NPAR!). Hope this helps, Jeroen http://pd-value.com [email protected]<mailto:[email protected]> @PD_value +31 6 23118438 -- More value out of your data! Op 20-02-16 om 21:01 schreef Mark Sale: Matts, Thanks for your insights. But, the issue isn't the post hoc values. With the mixture model the OMEGA on the intercept is huge (680), and the entire population is in the low intercept value group (Intercept = -11). Then to accommodate the patients with frequent AEs, it assigns a (post hoc) ETA of +15, giving an individual value for intercept of 6 (and a probability of the AE of ~1, as it should. My question is whey does it refuse to simply put those 8% of the patients in a sub population with intercept = 6, ETA=0. rather than saying the expected value is -11, with ETA = +15. Even when I fix the fractions in the subpopulations for the observed values, and fix OMEGA to a small, reasonable value, and fix the intercept values for the 3 populations to reasonable values it will still do this. The only thing that has worked is to assign each subject to the apparent population in the data set. Mark Mark Sale M.D. Vice President, Modeling and Simulation Nuventra, Inc. ™ 2525 Meridian Parkway, Suite 280 Research Triangle Park, NC 27713 Office (919)-973-0383 [email protected]<[email protected]> http://www.nuventra.com Empower your Pipeline CONFIDENTIALITY NOTICE The information in this transmittal (including attachments, if any) may be privileged and confidential and is intended only for the recipient(s) listed above. Any review, use, disclosure, distribution or copying of this transmittal, in any form, is prohibited except by or on behalf of the intended recipient(s). If you have received this transmittal in error, please notify me immediately by reply email and destroy all copies of the transmittal. ________________________________ From: Matts Kågedal <[email protected]><mailto:[email protected]> Sent: Saturday, February 20, 2016 2:44 PM To: Mark Sale Cc: [email protected]<mailto:[email protected]> Subject: Re: [NMusers] Mixture model with logistic regression Hi Mark, The pattern you see in the posthocs could possibly be a shrinkage phenomenon. I.e. patients with AE most of the time will have the same ETA, while patients with no AE will have the same ETA and there will be a third group in between. If shrinkage is causing this, you should not expect any improvement with a mixture model. Before you reject your original model I would therefore also evaluate it by simulation and re-estimation. I think it is quite possible that you will retreive a similar pattern in the posthocs even when you simulate based on a normal distribution. Best, Matts Kågedal Pharmacometrics, Genentech. On Fri, Feb 19, 2016 at 2:30 PM, Mark Sale <[email protected]<mailto:[email protected]>> wrote: Has anyone every tried to use a mixture model with logistic regression? I have data on a AE in several hundred patients, measured multiple times (10-20 times per patient). Examining the data it is clear that, independent of drug concentration, there is very wide distribution of this AE, 68% of the patients never have the AE, 25% have it about 20% of the time and the rest have it pretty much continuously, regardless of drug concentration. (in ordinary logistic regression, just glm in R, there is also a nice concentration effect on the AE in addition). Running the usual logistic model, not surprisingly, I get a really big ETA on the intercept, with 68% of the people having ETA small negative, 25% ETA ~ 1 and 7% ETA ~ 10. No covariates seem particularly predictive of the post hoc ETA. I thought I could use a mixture model, with 3 modes, but it refused to do that, giving me essentially 0% in the 2nd and 3rd distribution, still with the really large OMEGA for the intercept. Even when I FIX the OMEGA to a reasonable number, I still get essentially no one in the 2nd and 3rd distribution. I tried fixing the fraction in the 2nd and 3rd distribution (and OMEGA), and it still gave me a very small difference in the intercept for the 2nd and 3rd populations. Is there an issue with using mixture models with logistic regression? I'm just using FOCE, Laplacian, without interaction, and LIKE. Any ideas? Mark Mark Sale M.D. Vice President, Modeling and Simulation Nuventra, Inc. ™ 2525 Meridian Parkway, Suite 280 Research Triangle Park, NC 27713 Office (919)-973-0383<tel:%28919%29-973-0383> [email http://[email protected] http://www.nuventra.com