Testing moderation effect with TSCS(Pooled Time Series Cross Section) data

[Jin Lee]

Hi!
(I'm an international student, so maybe my English will be so awkward.. please understand it..!)

I want to test moderation effect with TSCS(similar with panel data structure) data.
In my data, There're 35 country's 12 years observations. So, total observation numbers are 420(=35*12).
I have 2 control variables, 2 independent variables, 3 moderate variables, and 6 interaction terms(=2*3).

y c1 c2 x1 x2 m1 m2 m3 (x1m1 x1m2 x1m3 x2m1 x2m2 x2m3)

According to the search, Moderation effect can be analyzed by hierarchical regression method and to confirm the existence of moderation effect, the researcher should check the variation of R2.

However I use a panel data, so I can't get an R2...
If I could use pooled OLS, it is easily solved, but mine isn't...
In my data, there are heteroscedasticity and autocorrelation problems.

If I just focus on my TSCS data, I think I can use a -xtreg or -xtgls (and I heard that large N small T is much better to use -xtreg!)
BUT... my research question is focused on to test the moderation effect.


How can I test the moderation effect with TSCS data? Is it possible..?
I feel like I'm in chaos..!


I need a help from an expert... Help me, please!

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful reply by following the FAQ on asking questions - provide Stata code in code delimiters, Stata output, and sample data using dataex.

xtreg with clustered or robust standard errors takes care of heteroscedasticity and autocorrelation.

Not everyone agrees that change in R2 is the right way to think about moderation. First, you need to read about factor variables in Stata. x1 moderating x2 can be written in the regression equation as x1#x2. The significance of coefficient on the interaction tells you whether the moderation is statistically significant. Using the factor notation also means that the margins command will work with the interaction nicely.