I remember seeing a github repo for this a while ago. Might have been this one, but I'm not sure: github.com/AllanCameron...

If you use git a lot, use Python, or need AI assistants: yes. Otherwise: no, stick with RStudio - it's still great at what it does and keeps getting better.

Looks really interesting! I might use this as a case study when I teach statistics for biologist next year.

Thanks Daniel! Very happy to hear it!

It's the variance of the _estimator_ that matters. This is still true if you estimate higher moments like skewness and kurtosis. The variance of those estimators will be related to the sixth and eight moments of the measurements, but it's still a variance.

strsplit("this/is/my/path", "/")[[1]]

Not until all countries adopt the YYYY-MM-DD date standard!

It should still be using right censoring I think! The survival package is extremely powerful, but the documentation isn't always easy to follow. I've sometimes compared the results to manual calculations in toy examples just to make sure that it was using the model I thought it was.

The issue could be that for data with consecutive time periods, Surv expects three variables: start time, stop time, event indicator. So Surv(Days_Lag, Days, SurvivalDepressionTotal_Diff) in your case. You may need to remove the rows where Days_Lag is NA as well.

Depends on what your data looks like. Do you have consecutive time periods for each individual?

The number of records is much larger than the number of individuals (n), which seems to indicate that you have multiple rows per individual in your dataset. But it's not because of time-varying covariates or multiple events? You're doing something beyond a standard Cox regression, surely? :)

What model are you using? This looks like the output from a Cox regression with time-varying covariates. Correct? Can patients have more than one event?

Yes! Right assignment with pipes *chef's kiss*

Highly recommended! It solves the problems you're having and speeds up the installation of new packages.

Have you tried R2U? github.com/eddelbuettel...

I'm extremely biased, but I think Modern Statistics with R is a great choice 😁
www.modernstatisticswithr.com