Message/Author 


I am estimating multilevel, weighted, continuous time survival models and the predictor variables of interest is on the WITHIN level. However, I am unable to create plots for each level of the variable. Is this possible in the current release? Thanks! 


These plots are part of the PLOT2 setting. When I run Example 9.18, I get survival plots. Perhaps these are not what you want. 


Thanks for the quick reply. They are part of PLOT2. However, when I attempt to specify the 'Estimated log cumulative curve' plots, I am able to use any of the plots in the 'Simple Curves' and 'Curves to Compare', but the 'Curves for specific covariate values' causes the program to crash. 


Please send the input, data, and your license number to support@statmodel.com so we can look into this. 


Hello, I'm running a two level continuous time survival analysis with Mplus Version 6.11 as shown in example 9.18, specifying the PLOT2 option. After selecting "estimated survival curve" under the "curves for specific covariate values" option, I am only able to select my time to event information variable t under "select the variable to be plotted". None of my covariates shows up here. What do I have to do in order to select any of my covariates? Thanks in advance for your help! 


The adjusted estimated means plot is not available when numerical integration is used. Example 9.18 requires numerical integration. 


Hi, I am also trying to build up on example 9.18 and have two questions on that: 1) I like to calculate U^2 (Hauser 1978) and wonder how to modifiy the Mplus code of example 9.18 to get the log likelihood value of the null model. 2) How do I integrate timevarying covariatess on both levels in this analysis? I would greatly appreciate your help. 


1) The null model should be this (L0) MODEL: %WITHIN% t ON x@0; %BETWEEN% t ON w@0; U2=1L1/L0 2) See Section 3.3 in http://statmodel.com/download/SurvivalJSM3.pdf or section 4 in http://statmodel.com/download/lilyFinalReportV6.pdf You will essentially need to break the survival variable into as many survival variables as there are time periods in which the covariate changes. 


Thanks for your feedback, Tihomir. Is it possible to extend the discrete survival model (example 6.20/6.21) to a multilevel model? If so, I would highly appreciate if you could give me a hint how an exemplary Mplus code would look like. Would it be possible to include timevarying as well as timeinvariant covariates on both, level 1 and 2? Thanks. 


This can be done by simply letting f of 6.20 have betweenlevel variation, or also letting the u's have random intercepts that vary on between. So on Between you say fb BY u1u5, and possible also u1u5 to free the random intercept variances. You can have tvc's on both levels. 

Stine Hoj posted on Tuesday, March 26, 2019  1:48 pm



Hello, I'd like to follow up on Tihomir's response above (Nov 29 2011) regarding integration of timevarying covariates into continuous time survival analysis. I have read Section 3.3 of the 2006 JSM paper where the survival variable (T) is "broken" into as many survival variables as there are time periods in which the covariate changes (Tk). The example in the paper results in a wideformat record for each participant in the order T1,...,Tk, Y1,...,Yk, d1,...,dk (where Y is the covariate and d the censoring indicator). However, I cannot figure out how to analyse the resulting data structure in Mplus. Could you please point me to any appropriate input examples or provide further guidance as to how this is achieved? Thanks in advance. 


I think you want this: T1Tk PON Y1Yk (1); or equivalently T1 on Y1 (1); T2 on Y2 (1); .... etc You actually don't need to hold these equal across the intervals. The scripts of the paper are available here http://statmodel.com/download/survival16.zip 

Stine Hoj posted on Tuesday, March 26, 2019  2:42 pm



Hi Tihomir, thanks so much for your prompt response. This may be silly, but it seems to me that we are analysing multiple time periods nested within individuals  is this taken into account in the model, or do we need to introduce a twolevel structure? Thanks again. 


It is multivariate (instead of multilevel  but yeah same thing). Take a look at the example in the zip file particularly T on the random intercept (the slope is slightly trickier). There is actually a subtle difference between multivariate and mutilevel and multivariate is the one you want here. In multilevel settings T0 T1 T2 T3 would have the same baseline (not good) but not with multivariate which does not assume the same baseline. 

Back to top 