Threshold models in Mark

[Deborah]

Hello all,

I am a PhD student working on senescence in long lived species, I am trying to fit threshold models on survival with Mark. My aim is to detect the ages at maturity and at onset of senescence but mainly the trends in increase and decrease of survival with age.

I thus use 2 thresholds of age each varying within ten values, and this gives me 3 phases of the curve of survival estimates versus age. On each of these phases I fit linear, constant, quadratic, weibull and gompertz models and I intend to find which combination of models and thresholds explain the best my data. Following me?

I managed to do this by constraining the design matrix (one column intercept phase1, then trend phase 1, intercept phase 2 then trend phase 2...). But my problem is that the results on each phase are independant, giving a non continous very ugly curve that is not right and unusable. Even if I get a global AIC value that allows me to compare with other models.

My question is then, do you have any idea on how to constrain in the design matrix the survival estimates/curve to be continous? e.g. telling a phase to start from the last value of the previous phase, or creating subsets or anything that would be right ?

I hope you have some ideas that could help me, and also that this post could help other people.

thank you in advance,
Deborah