How to correct autocorrelation in a multiple regression?

[Sunniva Davidson]

Hello everyone on statalist! Hope you had a great weekend.

I am working on a project and have run into a few obstacles.

The purpose of my project is to conduct a multiple regression analysis, with the stock price of an airline company as a dependent variable. For independent variables, I have used Oil Price, Google Trends activity, average temperature deviation in the country where the airline has most of its departures and USD/NOK exchange rate. These variables are all daily variables, and I have included all dates, even though some of the above-mentioned variables do not have values accounting for all dates, such as the weekends for oil price for example.

I have a question regarding the presence of autocorrelation in my regression. How do I account for serial correlation in the error terms in a multiple linear regression? I have tried the Cochrane-Orcutt procedure, using the prais command with the 'corc' option included at the end, but that causes R^2 to dramatically decrease and it rendered all my variables insignificant. What would you recommend for dealing with autocorrelation?

Thank you in advance!

[Carlo Lazzaro]

Sunniva:
are you referring to:
-regress- or -xtreg-?
autocorrelation or across panels correlation?

[Sunniva Davidson]

Carlo:
I am referring to -regress- and autocorrelation.

Thank you!

Kind regards,
Sunniva

[Sven-Kristjan Bormann]

Do you use the stock prices or the log-returns or first-differences of them as your dependent variable?
You should check if your dependent variable has a unit-root. You can use for example the Augmented Dickey-Fuller test via the - dfuller - command. If the test is significant, then you cannot use the original stock prices as your dependent variable in a normal regression. In fact, I would test all independent variables if they follow a unit-root process.
For further explanation, you can check any introductory text to time-series analysis or the Stata TS-manual.