Loop with missing values

[Ikn Jansen]

Dear all,

I want to estimate betas of the Fama French Model for every year. I have got weekly returns and I am using the following loop:


foreach variable of varlist r_share* {
sum `variable' if year == 1991
if r(N) >= 52 {
scalar lagged = floor(4*(r(N)/100)^(2/9))
newey `variable' r_sp500 smb hml if year == 1991, lag(`= lagged')
}
else {
display "Not enough observations"
}
matrix matrix_1991 = nullmat(matrix_1991) \ e(b)
}


Variables:
r_share = return share
smb = Fama French factor
hml = Fama French factor
r_sp500 = return s&p500

The loop is working fine there is just one problem and I am not able to solve this. As I need all the estimated betas in a matrix afterwards (matrix_1991) and I have got missing observations (for the Y-values) how can I tell stata to save "." in the matrix when there are "Not enough observations"?

So if I have got enought observations Stata should do the newey west regression and if not it should just store . for all coefficients.

Hopefully somebody can help me out!

BR

Ikn

[Clyde Schechter]

Well, instead of (or in addition to) -display "not enough observations"- run

matrix junk = (.)
matrix matrix_1991 = nullmat(matrix_1991) \ junk

[Ikn Jansen]

Thank you Clyde for your quick response. At least my loop is calculating now but unfortunately I receive a comformability error.

My code looks as follows:

foreach variable of varlist r_share* {
sum `variable' if year == 1991
if r(N) >= 52 {
scalar lagged = floor(4*(r(N)/100)^(2/9))
newey `variable' r_sp500 smb hml if year == 1991, lag(`= lagged')
}
else {
matrix junk = (.)
matrix matrix_1991 = nullmat(matrix_1991) \ junk
}
matrix matrix_1991 = nullmat(matrix_1991) \ e(b)
}

I guess Stata or at least me can't fill the new matrix (newey west matrix + junk matrix) with only newey west coefficients or is there a solution for this?

[Clyde Schechter]

Right, sorry. The matrix junk can't just be a single missing value because it won't conform to the shape of matrix_1991. So you have to make it have as many missing values in it as you have coefficients in the e(b) matrices when -newey- runs successfully. So, let's say that the e(b) matrices all have 4 colulmns, then -matrix junk = (., ., ., .)- is needed.

[Ikn Jansen]

Wow thank you Clyde! You really helped me out. Really appreciate that!