Newey-West Standard Errors

[William Wijkmark]

I’m very new to Stata and statistical testing in general, but I have grown very fond of it.

I’m currently trying to analyze calendar anomalies on stock data, I have approximately 3700 observations of daily returns of which I have created dummy variables for each month of the year.

I initially made a simple linear regression containing all dummies, but later decided to go for a regression with Newey West standard errors to avoid autocorrelation and heteroscedasticity. I have so far applied lag(254) to represent the number of open stock market days per year, and hence one cycle of testing. Furthermore, I decided to remove the constant to avoid the any variables being omitted.

Command: newey stock_return dummy_jan dummy_feb dummy_mars dummy_apr dummy_maj dummy_juni dummy_juli dummy_aug dummy_sept dummy_oct dummy_nov dummy_dec, lag(254) noconstant

What are your thoughts on this reasoning and processes? Thereafter I have tried to conduct a Breusch-Godfrey and Breusch-Pagan to evaluate the presence of autocorrelation and heteroscedasticity after the Newey regression, but seemingly this does not work. Any recommendations are kindly accepted.

If I have placed this post in the wrong spot, please forgive me as this is my first post on the forum.

[William Lisowski]

Welcome to Statalist.

You have accidentally posted your topic in Statalist's Mata Forum, which is used for discussions of Stata's Mata language, which is different than Stata's command language, and different than Stata's matrix commands. Your question will see a more appropriate, and much larger audience if you post it in Statalist's General Forum.

Also, if you have not already done so, take a look at the Statalist FAQ linked to at the top of this page for posting guidelines and suggestions. You will see that "seemingly this does not work" is not helpful, since there are countless ways in which something that should work can be accidentally made to fail.