Feature to visually inspect start parameters for model including covariates

[mptend]

Currently start.params is defined when the model is estimated. It would make initializing parameters easier to include them with the clvdata object or covariates object and be able to plot with the start.params and see a rough model fit before running the estimation. When I used to run these models in excel, visual inspection was the main way to ensure starting parameters led to a close to optimal solution.

[mmeierer]

Thanks for your comment. It would be interesting to learn more about this.

Did you run into problems when estimating latent attrition models with CLVTools? The estimation routines implemented in CLVTools should be quite reliable. I am aware that this did not necessarily apply to earlier implementations in other software.

If your experience differs, we would greatly appreciate if you could provide a reproducible example to illustrate this point. Also, a initial implementation how you would address this would be helpful.

[mptend]

I am having a hard time getting the KKT conditions to be TRUE. I have the latest from CRAN but I can try with the latest here on github and report back.

[mmeierer]

Getting KKT right can be challenging at times. Often re-estimating the model with the final parameters of the previous estimation run might work. Also switching the optimizer can help.

However, please be aware that validation with a test sample is as important. It provides key information on the applicability of a model for a real-world usage. Often practitioners focus on this metric.