Performing regression in stata

[Angelo Piazza]

Hello everyone,
I have to start a regression analysis in Stata in order to complete my master's degree thesis in Corporate Finance.
I have analyzed a dataset (from 2005 to 2018) of 300 Italian-listed family firms concerning implementing sustainable policies within the companies.
Now I have created several charts with my dataset and the result was that starting from 2014 in the automotive sector and in the retail (I have also other industries but just these 2 appeared to be relevant) one my ROE, ROI and M/B are higher in firms that implemented sustainable policies (through sustainable reports) with respect to the ones who didn't. However, I would like to study now if this can be also demonstrated with a statistical approach with a basic linear regression and verify the significance of the dummy variables (1 if the company published sust report 0 if it doesn't). What is the best way to compute it?
1) Do I have to put all the datasets from Excel concerning Automotive and Retail together or I have to create 2 different datasets?
2) If in my "financial" analysis I observe that starting from 2014 the metrics are higher, I have to insert the dataset starting from 2014 or I can upload the whole dataset
I would be very grateful if anyone could help.
Thank u so much
Angelo

[Carlo Lazzaro]

Angelo:
welcome to this forum.
I'm under the impression that, in order to be successful, you should consider studying any decent textbook on panel data econometrics.
In addition, please read and act on the FAQ in order to increase you chances of getting more helpful replies.
That said:
1) I assume that you're dealing with a N>T panel dataset with a continuous regressand (Return on Investement or Return on Equity);
2 It's really hard to believe that you can perform a simple panel data regression with one predictor only. Even assuming that it reaches statistical significance it would be not informative; at all
3) I think that you should -append- both the dataset and include -i.industry- (ie, Automotive=0; Retail=1) as a categorical predictor.
4) I would consider the whole timespan (2005-2014) without and data mae-up;
As a sidelight, the seeming lack of supervisors's support on quantitative matters (if she/he is to blame) always puzzle me.