I cant set maxvar higher

[Anton Kusk]

. set maxvar 20000
no; dataset in memory has changed since last saved

What to do?
I have saved plenty

I have tried
. set maxvar 20000, clear
no; dataset in memory has changed since last saved

[Øyvind Snilsberg]

type "clear all"

[Anton Kusk]

the i clear my entire data set, what does that solve?

[Leonardo Guizzetti]

To elaborate a bit, if you want to change -maxvar-, it must be done before working with data.

[Anton Kusk]

Okay, thanks for the answers. So when i get the error message below, which i guess stems from my dataset being too big; there is nothing to do about all the variables i have created and the fact that i merged two datasets to begin with?

. regress ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro i.pair_id
maxvar too small
You have attempted to use an interaction with too many levels or attempted to fit a model with too
many variables. You need to increase maxvar; it is currently 2048. Use set maxvar; see help
maxvar.

If you are using factor variables and included an interaction that has lots of missing cells, try
set emptycells drop to reduce the required matrix size; see help set emptycells.

If you are using factor variables, you might have accidentally treated a continuous variable as a
categorical, resulting in lots of categories. Use the c. operator on such variables.

[FernandoRios]

Hi Anton
I think you are trying to FIT a model by explicitly adding a large set of dummies (pair_id).
That is in general a bad idea. MOstly because you may not be interested in those coefficients.
Your best approach would be doing something like the following:

areg ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro, abs(pair_id)

[Leonardo Guizzetti]

Okay, thanks for the answers. So when i get the error message below, which i guess stems from my dataset being too big; there is nothing to do about all the variables i have created and the fact that i merged two datasets to begin with?

. regress ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro i.pair_id
maxvar too small
You have attempted to use an interaction with too many levels or attempted to fit a model with too
many variables. You need to increase maxvar; it is currently 2048. Use set maxvar; see help
maxvar.

Fernando has given you good advice on how to handle estimation. Your error suggests that you are using Stata BE (see -help memory-) because the default, and maximum, maxvar is 2,048 variables. if that's true, then no matter what you do, you cannot exceed this limit with your current flavour of Stata.

[Anton Kusk]

Yeah okay, but this means that colony, comlang and ln_dist is omitted which makes sense if they are absorbed by the pard_id fixed effects, but my begge_euro whcih means both_euro is suddenly a negative coefficient, which makes no sense for me

[Anton Kusk]

I also need year fixed effect, which i tried to add by; regress ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro exp_time_fe
But i get the maxvar error here as well

[FernandoRios]

Hi Anton
The advice I provide earlier remains. You are trying to estimate a model with too many variables, mostly because of interactions.
If some results come with unexpected signs it could be an indication of some other type of problem, but that requires further exploration and a better understanding of the data and the topic.

Simply said. If you are trying to add fixed effects, do not try using "i." that is a good idea only when you have a few categories. If you want to add time and pair ID fixed effects you can use the user-written command "reghdfe" (ssc install reghdfe).

Beyond that advice, I think you need to revise what are you trying to control, how is that being affected by the fixed effects or the time trend.
HTH

[Anton Kusk]

First of all, thank u very much Fernando!
I was able to run the exp_time fixed effects this way, but i am still lost on how to add the imp_time fixed effects and pair_id fixed effect to the same command
reghdfe ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro, abs (exp_time)

I am very new to STATA, and i am so close at having this model up and running sufficiently enough to write my BA

[Øyvind Snilsberg]

reghdfe can absorb multiple fixed effects,

reghdfe ln_trade ln_distw ln_gdp_both colony comlang_off begge_euro, abs(exp_time imp_time pair_id)

[Anton Kusk]

Ahh okay. THANK U!

And i guess the reason why distance (distw) and the other variables are omitted is simply because they are now absorbed by the fixed effects of either pair or exporter time or importer time?

------------------------------------------------------------------------------
ln_trade | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
ln_distw | 0 (omitted)
ln_gdp_both | 0 (omitted)
colony | 0 (omitted)
comlang_off | 0 (omitted)
begge_euro | -.2241433 .171 -1.31 0.190 -.5593016 .111015
eu_o | 0 (omitted)
eu_d | 0 (omitted)
_cons | 16.10682 .001514 1.1e+04 0.000 16.10385 16.10978
------------------------------------------------------------------------------

[Øyvind Snilsberg]

And i guess the reason why distance (distw) and the other variables are omitted is simply because they are now absorbed by the fixed effects of either pair or exporter time or importer time?

Correct.

[Anton Kusk]

Thank u!
(maybe this isent Stata related, but for interpretation, do i have to remove the omitted variables from the model and run it again or?)