Calculating statistical power

[Lieke Breure]

Dear, for my thesis I run a multiple regression analysis, for which there is one variable of interest. I ran the following regression:
reghdfe LN_Dealvalue LN_REVoverCOGS Acquiredstakecategorical dummy_international dummysameindustry LN_revenue LN_EBITDA LN_targetage LN_employee, absorb ( Targetcountrycode Acquirorcountrycode yearcategorical)
My results can be viewed in the attachment.

Now, my thesis supervisor says that I need to calculate the statistical power. So what should be the sample size to be able to say something on the coefficient of LN_REVoverCOGS?

Can someone help me in how to do that?

Thanks in advance!

[Carlo Lazzaro]

Lieke:
see Example 2: Sample size for testing a subset of coefficients, -power rsquared- entry, Stata .pdf manual.

[Lieke Breure]

Thank Carlo!
Just to be sure, the reduced model would be the one containing only the control variables, excluding LN_REVoverCOGS?
Resulting in the command
power rsquared 0.7975 0.7995 , ncontrol(7)


with:
alpha = 0.0500
power = 0.8000
delta = 0.0100
R2_R = 0.7975
R2_F = 0.7995
R2_diff = 0.0020
ncontrol = 7
ntested = 1

Estimated sample size:

N = 789

[Carlo Lazzaro]

Lieke:
your code seems to miss the -ntestedl()- portion.
In addition: where does R_sq 0.7975 come from?

[Lieke Breure]

Hi Carlo, Ntested is normally always 1 right?

0.7975 is the Rsquared of the regression
reghdfe LN_Dealvalue Acquiredstakecategorical dummy_international dummysameindustry LN_revenue LN_EBITDA LN_targetage LN_employee, absorb ( Targetcountrycode Acquirorcountrycode yearcategorical)
--> Thus excluding LN_REVoverCOGS

Rsquared of 0.7995 is the Rsquared when you include LN_REVoverCOGS

[Carlo Lazzaro]

LIeke:
thanks for clarifying.
I'd go (I assume that you are interested in the sample size necessary to test the null by adding 1 more predictor (that is, from 7 to 8):

power rsquared 0.7975 0.7995 , ncontrol(7) ntested(1)