Estimating the non-zero asymptote in logarithmic day from individual level data

[Robert Shaw]

Dear Statalisters,

This may be more of a statistical question than a programming one. I have panel data for several individuals separated into n=8 groups. Each participant has multiple measures of a biological parameter which decreases over time. The rate of decay appears to be logarithmic with a non-zero asymptote, which is likely different between different groups. Each individual's timepoint measurements have a 'nominal label' (eg a the Day 100 visit), but in actuality this visit will have occurred over a number of different days from starting the study (eg some will have had their visit on day 91, some on day 103). How might I use the individual level data including the variation in real timepoint to estimate the non-zero asymptote per group? Is heading down the non-linear mixed effects model the correct road to walk down?

Kind regards
Robert Shaw

[Nick Cox]

Question says with a non-zero asymptote, text says without. I guess with. Do you mean y = a + y_0 exp(-b t) where a varies?

[Robert Shaw]

Yes, thank you for picking up on that error. A criticial one! I have edited post #1. Yes precisely how to estimate `a' in y = a + y_0 exp (-b t)
Another typo is in the title of the thread. Day should be Decay, but it is too late to edit the post.

[Nick Cox]

nl may help.