Invalid Syntax

[mark zuckerberg]

We wanted to create an do-file for a coefficient-estimation for the previos 36 months for each fund and then calculate a 12-month ahead estimate for alpha , but when we want to apply this do file to our dataset we always get the error message 'invalid syntax' r(198), but stata does not specify where our problem is.


//load dataset




gen SPret = benchmark - rf


//generate variables for results with momentum

gen alpha = .


gen beta_SPret = .

gen beta_smb = .

gen beta_hml = .

gen beta_mom = .

gen r2 = .

//generate variables for results (without momentum)

gen alpha_wom = .

gen beta_SPret_wom = .

gen beta_smb_wom = .

gen beta_hml_wom = .

gen r2_wom = .

//begin loop



//create group identifier with variable code

egen id = group(name)

sort name year month

sum id

local max `r(max)'


forvalues i = 1/`max' {


foreach id == `i' {

local y = min(calmt) //create local min of calmt to know where to start
local z = max(calmt) //create local max to identify ending point




di in red "`c(current_time)': @id: `i'" //show loop results


sum id if id == `i' // Calculate the sum of observations for each funds


if `r(N)' > 47 { // There have to be at least 47 months calmt observations, because the coefficients are calculated over a period of 36 months and then an alpha is calculated 12 months ahead



rolling _b, start(y) window(36) end(z-12) saving(betas_m) stepsize(1) : reg mgret SPret smb hml mom



replace alpha = _b[_cons] if id == `i'


replace beta_SPret = _b[SPret] if id == `i'


replace beta_smb = _b[smb] if id == `i'


replace beta_hml = _b[hml] if id == `i'


replace beta_mom = _b[mom] if id == `i'


replace r2 = e(r2) if id == `i'
}
}
}

di in red "`c(current_time)': @id: `i'" //show loop results

sort name year month

sum id

local max `r(max)'


forvalues i = 1/`max' {

foreach id == `i' {


local y = min(calmt) //create local min of calmt to know where to start
local z = max(calmt) //create local max to identify ending point

di in red "`c(current_time)': @id: `i'" //show loop results

if `r(N)' > 47 {
//regression without momentum
rolling _b, start(y) window(36) end(z-12) saving(betas_wom) stepsize(1): reg mgret SPret smb hml




replace alpha = _b[_cons] if id == `i'


replace beta_SPret = _b[SPret] if id == `i'



replace beta_smb = _b[smb] if id == `i'



replace beta_hml = _b[hml] if id == `i'



replace r2_wom = e(r2_wom) if id == `i'

}
}


}


}

[Nick Cox]

This is a very old-fashioned piece of code. The forum contains many posts on community-contributed commands that will do much for this for you.

That aside, you should be able to tell us more precisely where your code fails. To my eye, this looks OK

//load dataset
gen SPret = benchmark - rf

//generate variables for results with momentum
gen alpha = .
gen beta_SPret = .
gen beta_smb = .
gen beta_hml = .
gen beta_mom = .
gen r2 = .

//generate variables for results (without momentum)
gen alpha_wom = .
gen beta_SPret_wom = .
gen beta_smb_wom = .
gen beta_hml_wom = .
gen r2_wom = .

//begin loop

//create group identifier with variable code

egen id = group(name)

sort name year month

sum id

local max `r(max)'

forvalues i = 1/`max' {

But the next three statements would all fail.

foreach id == `i' {

That is a long way from any legal variant on foreach. But you are already in a loop over your distinct identifiers. So, no inner loop is needed here any way.

local y = min(calmt) //create local min of calmt to know where to start
local z = max(calmt) //create local max to identify ending point

Those are illegal too. min() and max() as functions won't work on one argument. Nor do they know to operate differently within a loop according to the current loop value. Perhaps you meant to do something more like this:

su calmt if id == `i', meanonly
local min = r(min)
local max = r(max)

Also, looking ahead

replace r2_wom = e(r2_wom) if id == `i'

That's not going to work as you wish. regress doesn't leave e(r2_wom) in its wake.

I'll stop there. Someone is mixing borrowed and guessed code. Borrowing often works, but guessing rarely.

More generally, "mark zuckerberg" just possibly could be your real name, but we do flag that we request full real names on Statalist.