Can you describe in detail what you are referring to exactly when you say "fixed effects"?
I am assuming you ran a regression of some sort with no random effects, but I am unsure exactly what you did.
Hi all,
Let me shortly introduce my research: I am looking at whether companies outperform each other in investments. To do this I observe them at multiple investments, and calculate fixed effects. Afterwards, I tested the fixed effects to confirm whether at least one company outperforms other companies.
The next step is to look at the distribution of the fixed effects and test them for normality(suggested by other studies). I tested for normality with the Jarque-Bera test, the outcome is that the distribution of fixed effects is not normal. However, I fail to see what this means.
My question: how do I interpret the fixed effects being not normal?
Kind regards,
Daniel
Can you describe in detail what you are referring to exactly when you say "fixed effects"?
I am assuming you ran a regression of some sort with no random effects, but I am unsure exactly what you did.
Stop cowardice, ban guns!
daniel_smit (06-09-2017)
First of all, thank you for the response. To specify a little more what I did:
1. Running fixed regression (a dummy for each company)
2. Testing the following hypothesis trough a partial f-test: Company_1=Company_2=.....=Company_n. In words: companies add the same levels of value to their investments.
3. Outcome: companies add different levels of value to their investments.
4. Testing the distribution of the company fixed effects with a Jarque-Bera test.
5. Outcome: distribution is not normal.
6. What does this mean exactly?
Does this make it more clear?
Regards,
Daniel
Sounds like you may be thinking of running perhaps an ANOVA or simple regression model, yes? If so, the outcome variable does not need to be normal, it is the error terms in your model that need to be normal and independent.
Stop cowardice, ban guns!
Tweet |