I have a data set about military interventions from different countries in civil wars. It contains a list of civil wars with 4 possible outcomes.

1. US intervention
2. Soviet Intervention (its a cold war data set) 
3. Intervention by 'other'
4. No outside intervention

The data set contains, per war, a lot of variables on the nation that is suffering the war (GDP, political system, location, etc)

I am running Logit regressions to find out which factors influence intervention from one of the 2 major cold war factions.
(For example, is the 'democracy score' significant for the United States, or does is the fact that the country is fighting a war of independence statistically significant for the Soviet Union?)

However, I was wondering that if I analyzed, for example, US interventions whether or not I should include / exclude Non-US interventions.

When looking at previous research, I see researchers do both. In one article the researcher excluded all Non-US interventions, reasoning that he was only interested factors that influenced US interventions.

Another article ran the regressions on a single data set that included all nations. Neither really went into the 'why' of their decision, leaving me to wonder whether their results were distorted.

Wouldn't excluding 'NON-US' interventions create 'false positives'? (As including cases where US intervention did not happen might reveal them to be just a fluke). In many cases the N is somewhat small (below 120), so I am a bid paranoid about things that might distort the results.

When running regressions I've also made sure to include 'negative' results in order to get a more complete / realistic model, and I would assume that would apply in this situation as well, am I missing something?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.