search

imperialCHEPI / healthgps

Global Health Policy Simulation model (Health-GPS)
https://imperialchepi.github.io/healthgps/
BSD 3-Clause "New" or "Revised" License
4 stars 1 forks source link

Null value in nutrients under intervention #281

Closed jzhu20 closed 5 months ago

jzhu20 commented 9 months ago

Null values are found in foodcarbohydrate, foodfat, foodprotein, foodsodium, energy intake, weight, bmi, under intervention, in some specific years. I'm not sure if it's an issue. For example,

Under policy scenario 1

Female

Null values in 2037, 2039, 2040

Male

Null values in 2030, 2036, 2050, 2051, 2055

Under policy scenario 2

Female

Null values in 2030, 2036, 2037, 2040, 2043

Male

Null values in 2028, 2029, 2051

jamesturner246 commented 9 months ago

Is this from the same simulation as #277?

jamesturner246 commented 9 months ago

Also, could you give a bit of info about where these NaN are appearing? Are they in the output csv file, or the json file?

Could you upload the file these appear?

jzhu20 commented 9 months ago

All are based on the same simulation. NaN appeared in output csv files. It can be found in this sharefolder. https://imperiallondon-my.sharepoint.com/:f:/r/personal/jjo11_ic_ac_uk/Documents/CHEPI/research/franco/Health-GPS_SHARED/Health-GPS_INDIA/Visualization?csf=1&web=1&e=RQOLaa For example, in HealthGPS_Result_S1.csv, lines 3630, 6056, 6461, etc.

jamesturner246 commented 9 months ago

Okay. I suggest we make a minimal example which produces these NaN results. Small a population and runtime as possible, with no other cruft to distract us. Can talk about it tomorrow.

jzhu20 commented 9 months ago

No NaN in France example results.

jamesturner246 commented 9 months ago

Great, thanks. Need to figure out if it's still happening with India data now.

jzhu20 commented 9 months ago

No NaN in results with India/France inputs and India diseases data.

jamesturner246 commented 9 months ago

Ah fantastic, thanks. Can this issue be considered resolved now? Or is there some other permutation of data which is problematic?

jzhu20 commented 9 months ago

Shall we wait until we get results using the new KH model? There might be problems there.

jzhu20 commented 8 months ago

I got "nan" in simulation results using KH model. @jamesturner246 Results can be found here. https://imperiallondon-my.sharepoint.com/:f:/r/personal/jzhu5_ic_ac_uk/Documents/Project/RSTL_India/Visualization/[Iteration18-India-fulllist](https://imperiallondon-my.sharepoint.com/:f:/r/personal/jzhu5_ic_ac_uk/Documents/Project/RSTL_India/Visualization/Iteration18-India-fulllist?csf=1&web=1&e=r5M0R9)?csf=1&web=1&e=r5M0R9

jamesturner246 commented 8 months ago

Great, thanks. link is dead. Assume it's still uploading?

jzhu20 commented 8 months ago

Sorry. @jamesturner246 Can you check with this link? I don't know what happened to be honest. https://imperiallondon-my.sharepoint.com/:f:/r/personal/jjo11_ic_ac_uk/Documents/CHEPI/research/franco/Health-GPS_SHARED/Health-GPS_INDIA/Visualization/Iteration18-India-fulllist?csf=1&web=1&e=INGICz

jamesturner246 commented 8 months ago

Thanks @jzhu20. That's better. I'll investigate. This is with the plain old example_new, right? Or some other config?

Edit: This is interesting, in array index 1, the NaNs seem to begin with one of the group of 92 year old males, and recur year after year (93 ,94, 95, ...) until removal. It's like a single 92 year old male (or a few) is throwing off the average for the rest of the group.

There's other instances where groups encounter NaN means, and then the age plus one group encounters NaNs in the following year, and so on...

I'll try to narrow this down, but would help to know the config used.

jamesturner246 commented 5 months ago

Done in https://github.com/imperialCHEPI/healthgps/pull/353 and https://github.com/imperialCHEPI/healthgps/pull/356.