0

so I currently trying to get an good regression for my gravity equation with sanction impact. At first I tried my luck with PPML, but the package refuses to estimate without the distance variable, but the variable is contradictory to my dyed-fixed Effects. So I was forced to look for an other package and found the FENmlm-Package, which also accepted my situation with a lot of zeros in the dataset so I was very happy until estimation in R printed the error message "The optimization algorithm did not converge, the results are not reliable." Thats pretty much my situation in this moment, but then I found out I could insert a part that's called precise.cluster, that would help with the converge, but until now I couldn't explain myself where I need to insert this part. Maybe someone can help me:) I found the package on this site: https://cran.r-project.org/web/packages/FENmlm/vignettes/FENmlm.html#232_a_note_on_precision Thanks for the help in advance

The code is: regression_devfeNmlm = feNmlm(trade.x ~ rta + arms | ex_yr + im_yr + ex_im , family = poisson, df_merge_ind5)

summary(regression_devfeNmlm)

Bobby X
  • 1
  • 1
  • 1
    Hi, welcome to StackOverflow. Can you provide a small dataset and accompanying code to reproduce the error, or a small dataset and your coding attempts so far to implement the `precise.cluster` part? That would make it much easier for people to understand what is going wrong. Some instructions to create a reproducible set can be found here; https://stackoverflow.com/a/5963610/5805670 – slamballais May 15 '21 at 10:20
  • @slamballais I am sorry I not so talented with this program so it's not possible for me to give you all an small dataset to reproduce the problem – Bobby X May 15 '21 at 16:51
  • @slamballais my question is just where I have to put the term precise.cluster – Bobby X May 15 '21 at 16:53
  • So `precision.cluster` seems to be an argument for `femlm`, which is the function that `FENmlm` runs. In other words, if you type `?femlm`, you can see exactly what `precision.cluster` does. However, it seems that the author moved on to a new package called `fixest`. You may want to look into that: https://cran.r-project.org/web/packages/fixest/vignettes/fixest_walkthrough.html – slamballais May 15 '21 at 18:13
  • year thanks I tried fixest a few weeks before and I think I had and problem or so. So I searched for other opportunities, but with my current dataset it's seems to be working quiet well. So @slamballais thanks very much for bringing it to my attention again!! The tip helped me a lot!! :) – Bobby X May 15 '21 at 22:10

0 Answers0