Sensitivity test

[Klaudia Lennerling]

Hello fellow stata users!

I am in the midst of writing my bachelor thesis and I'm doing a cross-sectional analysis on a specific country. I am required to do a robustness test or a sensitivity analysis and I don't know what would be the most appropriate one to do. I have in my regression model added control variables and I have also done a correlation matrix but I think some motivated sensitivity analysis is needed. Do you have any suggestions on what to do?

my main regression is

reg Ln(annual sales) gender age locality, r

where:
Ln(annual sales) is the logged annual sales of the different firms
gender is a dummy if the firm is owned by a female or not
age is a control variable for how old the firm is
locality is a control variable and is the size of the region the firm is operating in

This regression is made for two separate years as well.

If you have any questions let me know, I have tried to research sensitivity analysises that would be appropriate but can't find anything that could be motivated. Thank you for your help!

Klaudia

[Carlo Lazzaro]

Klaudia:
1) as per your -regress- code, you've already invoked -robust- standard errors to take heteroskedasticity into account. You may want to elaborate on this topic via presenting FGLS and WLs regression that are well covered in

https://www.stata.com/bookstore/microeconometrics-stata (Chapter 5);
2) did you check for autocorrelation of the residual too?
3) you may want to include both linear and squared term for -age- in the right-hand side of your regression equation:

c.age##c.age

to search for possible turning point;
4) -linktest- comes in handy to test if the functional form of your regressand in misspecified (points 4) and 3) are strictly linked).

The usual advice of planning all these (and posibly many others) anaytical steps with your professor/suprvisor(teacher obviously applies.

[Klaudia Lennerling]

Thank you very much, we'll try your suggestions

All the best,
Klaudia