Extrem RRR value in logit regression

[samaneh khaef]

Hi
I am running mlogit to for knowing which places immigrants choose to live initially.
my dependent variable is the type of cites and age, year of arrival, category of immigrants, their education, and country of origin
mlogit Citytype age i.cohort(year of arrival) i.education i.origin i.catry of migrants

But the RRR is extremely high for Nordic migrants and I don't know how to solve it

Rural areas is my reference category.

[Carlo Lazzaro]

Samaneh:
I would start with revising that data related to Nordic migrants, just to rule out any nasty data entry.

[samaneh khaef]

Thanks. But can you explain a bit more? do you mean to remove all Nordic or what ?

[Carlo Lazzaro]

Samaneh:
not quite.
I would double check for trivial errors in data entry as far as Nordic migrants are concerned.

[samaneh khaef]

would it be the reason that Origin and Catergoy may interact? when I remove any of these two variables the RRR become logical

[samaneh khaef]

well
I am not really good at this. how can I check for this error?

[Carlo Lazzaro]

Samaneh:
could you please share an excerpt/example of your dataset via -dataex-? Thanks.

[samaneh khaef]

This is a very small sample of data as I sadly I am not allowed to share the whole dataset

[Carlo Lazzaro]

Samaneh:
please use -dataex- (not screenshots) to share excerpt/example of your dataset (see the FAQ). Thanks.

[samaneh khaef]

Type Category age year of migration eductaion origin Large Nordic short visit 34 2008 High Nordic Metropolitan EU short visit 56 2011 High Europe, excluding Nordic Metropolitan Family 56 2010 Low Asia Rural Student 45 2011 High Asia Medium Refugee 26 2009 Low Asia Large EU short visit 45 2010 High Europe, excluding Nordic Rural Family 65 Low South America Lareg Labor 44 2009 High African Medium EU short visit 56 2009 High Europe, excluding Nordic Small Nordic short visit 67 2011 High Nordic Rural Nordic short visit 34 2010 Medium Nordic Rural Nordic short visit 43 2010 High Nordic Large Labor 34 2008 Medium South America Metropolitan Refugee 45 Low Asia Small Family 27 2009 Low North America Rural Nordic short visit 29 2011 High Nordic Large Refugee 23 2008 Medium African Medium Student 55 High Oceania and Australia Medium Family 67 2008 Medium African Metropolitan Labor 54 2011 Low North America Metropolitan Labor 43 2010 Medium North America Medium Student 34 2011 High North America

All I can is to insert it as table as I am working from platform and am not allowed to extract any data.

[Carlo Lazzaro]

Samaneh:
as far as I can assess it, the issue you're facing might be due to -Nordic short visit- that seems to be repeated for -Nordic- origin only.
Sorry that I cannot be more helpful.

[samaneh khaef]

Sure Thanks! Yes it sounds to be true as when I insert only category or origin the result is okay. so do you have any suggestions on how can I have both variables without getting high rrr?

[samaneh khaef]

But that also can be the case for EU as I have one short visit EU and their origin is also EU. Right?

[Carlo Lazzaro]

Samaneh:
yes, it might be the case for EU, too.
However, -mlogit- works well when the different level of the categorical regressand differ enough within the predictors.
Hence, I would check (and compare, to have an idea of what's goin on with your data) -Category- and -Origin- for both -Nordic- and -EU (excliding Nordic)- countries.