summary with lean still produces a long vector obsRemoved

[zozotintin]

When using summary with option lean set to true, the field obsRemoved is not cleaned. Sometimes obsRemoved can be very long and close to order O(N), and I will have to clean it myself with model_summary$obsRemoved = NULL. I am not sure whether this value is used in calculating other statistics, but if not then perhaps this attribute can be cleaned to save more space.

[lrberge]

Thanks, it was an oversight... there are just too many objects!!! I've also added iv_residuals that I had forgotten to include.

[zozotintin]

I just found that in the IV regression object, the iv_first_stage and se_info still carries O(N) objects. Not sure whether you fixed the problem or not. I was using commit a1053d7f.

[zozotintin]

FYI, this is the code I used to trim down the IV model in commit a1053d7f

model_summary$iv_first_stage$fixef_id = NULL
model_summary$iv_first_stage$obsRemoved = NULL
model_summary$iv_first_stage$residuals = NULL
model_summary$iv_first_stage$sumFE = NULL
model_summary$iv_first_stage$fitted.values = NULL
model_summary$iv_first_stage$fitted.values_demean = NULL
model_summary$iv_first_stage$scores = NULL
model_summary$se_info$cluster = NULL
model_summary$summary_flags$call_env = NULL

[lrberge]

Thanks for the report, it should be fixed now.

[zozotintin]

I just found that in 0.8.3, the fitted.values_demean in iv_first_stage is still not cleaned.

[lrberge]

Gosh, I will never get rid of it.... Thanks for the clear report, I've updated just now. We're getting closer :-)