0

I apologize my knowledge in the field of statistics is limited as I am a newcomer to this area. I am trying to expand my understanding to concepts beyond simple linear regression and simple mixed effects models.

I have a situation that requires the application of tobit or tobin models to a dataset that has different left censored values depending on the lab that analyzed the blood sample.

I came across this example from survival package, https://cran.r-project.org/web/packages/survival/survival.pdf where the author describes the use of this model using tobin dataset

tobinfit <- survreg(Surv(durable, durable>0, type='left') ~ age + quant,
data=tobin, dist='gaussian')

My question is how do I expand this example to include random effects scenario where random intercepts represents the lab evaluating the sample and each lab has 2-3 different left censoring values. My covariates are age, education and BMI. I appreciate your patience and help in expanding my grasp on this subject.

Emily Fassbender
  • 145
  • 1
    Please see the question linked as a duplicate for an approach. Note that coding-specific questions are off-topic here. There is a statistical issue, in that it's hard to fit random effects to standard parametric survival models, and neither survreg() nor censReg() can do so directly. See section 5.5.3 of the R survival vignette for the problem with the "frailty" terms that work well for that purpose in Cox models. If you still have a question about statistical issues, please revise this question or ask a new one – EdM Apr 28 '23 at 15:31
  • @EdM, Thanks EdM – Emily Fassbender Apr 28 '23 at 17:02

0 Answers0