Newey-West regression

[Adrian Cernescu]

Hi, I am trying to run a Newey-West regression, however I do not quite understand how should I choose the number of lags for the regression, do I have to base the lags on the dependent variable or the independent variable. Also, my regression involves multiple independent variables, hence different lags for autocorrelation may be indicated.

CSAD = alpha + beta1(variable1) + beta2(variable2) + beta3(variable4) + residual

I know that to identify the lags of autocorrelation I would have to use

ac(variable)

, but since I have multiple variables, what should I do, should I just choose the largest lag value from all the independent and the dependent variable?

The data I am looking at is time series.

[Andrew Musau]

See #4 https://www.statalist.org/forums/for...lation-command

[Adrian Cernescu]

Thank you Andrew that helps a lot. Just to make sure, in my case I have data for 1152 observation so this will eventually mean that according to Stock H J. and Watson M. W. (2017). Introduction to Econometrics, 3rd Edition:

m=int(0.75∗T^1/3)
m=int(0.75∗1152^1/3)
m=7

So

m=7

will eventually be the lag value that I would use.
Since my regression contains multiple explanatory variables, do you reckon it is worth looking at

ac(variable)

for each of the dependent and independent variable and see if a value of m = 7, is capturing "all" lags for all variables, or is this rule of thumb enough?

[Andrew Musau]

The criterion is strictly a function of the number of observations \(T\). So m=7 is valid for all time-series regressions with 1152 observations, independent of the number of variables.