Keeping the residual and accounting for heteroscedasticity

[Raul Athwall]

I have two questions- I have time series data over 18 years for internet search terms- I have deasonalised this data on the day of the week and month dummies. How do I generate a new variable by keeping the residual for the terms on each day as a column in my data?
And also, I want to account for heteroscedasticity by dividing the residual by its standard deviation- is there a shortcut in doing this for all my variables in one go and having this as a new variable in a new column within my data?

[George Ford]

Why not use , robust to deal with heteroskedasticity?

regress y x , robust
predict efit, resid

[Raul Athwall]

Hi George,
What is the "efit" part of this? And also is there a way to do this for all my variables in one go as I have around 50?

[George Ford]

efit is the residual you wanted.

use auto, clear
foreach var in price mpg rep78 headroom trunk weight {
qui reg `var' foreign, robust
predict e_`var', residual
}

[George Ford]

You might want to look for autocorrelated errors as well.

[Raul Athwall]

Thanks for the help George,

With this I want to keep the residual and have this as a new column on data- e.g. for my search term recession, I want recession_robust as a new variable, with its new residual term. How would I go about doing this?

[George Ford]

That is exactly what post #4 does.