California-Planet-Search / radvel

General Toolkit for Modeling Radial Velocity Data
http://radvel.readthedocs.io
MIT License
58 stars 53 forks source link

Model comparison not saving results? #326

Closed zhexingli closed 4 years ago

zhexingli commented 4 years ago

Hello,

I'm not sure why but this one time model comparions doesn't seem to save the results? When I try to produce a radvel report, it says this:

WARNING: Could not find ic model comparison in ./HD140901_radvel/HD140901_radvel_radvel.stat. Please make sure that you have run radvel ic -t ic (or, e.g., radvel ic -t nplanets e trend jit gp) if you would like to include the model comparison table in the report.

Even though I did type the model comparison command in before generating the report:

radvel ic -t nplanets e trend jit -s HD140901_radvel.py

Why is this happening?

zhexingli commented 4 years ago

nvm, very stupid of me. This was only a one-planet fit, no need to include 'nplanets' in the model comparison. Leaving 'nplanets' out it's all good again.

zhexingli commented 4 years ago

Sry, I had to reopen it. I'm not sure, is there an issue with model comparison with 'nplanets'? Whenever I include 'nplanets' in the command "radvel ic -t xxx xxx -s [file]", doesn't matter if only doing number of planets comparison or together with other variables, my radvel returns error message when I do 'radvel report -s [file]':

WARNING: Could not find ic model comparison in ./HD134987_radvel/HD134987_radvel_radvel.stat. Please make sure that you have run radvel ic -t ic (or, e.g., radvel ic -t nplanets e trend jit gp) if you would like to include the model comparison table in the report.

I don't seem to get any warnings when doing report if I only do model comparison with other parameters such as 'e', 'trend', 'jit', or all together but without 'nplanets'. Whenever I got this warning message, of course, no model comparison is done and no such table is displayed in the final report.

I don't think this is the first time I'm having this issue. And this time I'm doing multi-planet fit.

bjfultn commented 4 years ago

OK, send the config file and data and I'll try to reproduce this behavior.

zhexingli commented 4 years ago

Thanks BJ. Here're the files packed in zip Files.zip

I have a feeling that the error message pops up at times when doing 'nplanets' comparison on data where all signals are significant. But sometimes it's hard to tell if they're all really significant...

bjfultn commented 4 years ago

This bug comes up because there are nan values in the particular ic comparison table. I've fixed this to allow for those bad values more gracefully in PR #388 which will be released as v1.4.2 shortly.

zhexingli commented 4 years ago

Awesome, thank you BJ!