Checking Interaction terms for significance using F-tests

[Jame Warrell]

Hi everyone I was wondering if I could please get some help with my regressions to get an F-test rather than a chi2 statistic, as my dissertation advisor has requested it in that format. Noting I have checked both the help and pdf file but none of the solution that proposed worked for me. I am aiming to interpret whether trust has differential effect across regions.

My code of interest is:

"ivregress 2sls log_GDP_pcc l.log_GDP_pcc log_educ log_ndg (log_GCF = l.log_GCF) log_general_trust Europe South_America North_America Asia Africa logged_trustx_Europe logged_trustx_South_America logged_trustx_North_America logged_trustx_Asia logged_trustx_Oceania logged_trustx_Africa i.Year, first robust

test log_general_trust-=logged_trustx_Europe=0"

Also if any one had any advise of how to demonstrate an F-test in pleasing way for my dissertation results that would be really helpful.

Thank you so much for taking the time to consider this.


yours sincerely,

James Warrell

P.s: Sorry if the solution is blaringly obvious.

[George Ford]

https://www.stata.com/support/faqs/statistics/chi-squared-and-f-distributions/

I wouldn't think that test line would work. Lots of excess characters.

What is trustx? Mean of the region?

To test whether mpg has different effects across rep78 using the F-test, something like this might work:

sysuse auto, clear
ivregress 2sls price mpg c.mpg#i.rep78 weight i.rep78 (displacement = gear_ratio), first robust
local df = `e(N)' - (`e(df_m)' + 1)
margins, dydx(mpg) over(rep78) post
test _b[mpg:1.rep78] = _b[mpg:2.rep78] , df(`df')
test _b[mpg:1.rep78] = _b[mpg:3.rep78] , df(`df')
test _b[mpg:1.rep78] = _b[mpg:4.rep78] , df(`df')
test _b[mpg:1.rep78] = _b[mpg:5.rep78] , df(`df')
test _b[mpg:2.rep78] = _b[mpg:3.rep78] , df(`df')
test _b[mpg:2.rep78] = _b[mpg:4.rep78] , df(`df')
test _b[mpg:2.rep78] = _b[mpg:5.rep78] , df(`df')
test _b[mpg:3.rep78] = _b[mpg:4.rep78] , df(`df')
test _b[mpg:3.rep78] = _b[mpg:5.rep78] , df(`df')
test _b[mpg:4.rep78] = _b[mpg:5.rep78] , df(`df')

[Jame Warrell]

trust x region i.e.:Asia is an interaction term between a region and the reported value in trust from the World Value Survey dataset. I'll try the code you've provided for the F-test with my dataset and thank you so much for the suggestion!

[George Ford]

I would use the factor approach in your regression model: Asia c.Asia#c.log_general_trust