lasso error message

[mathieu nacher]

hello, i have tried to test the lasso command using a dataset where i try to predict a continuous dependent variable using sets of categorical and continuous independent variables.
i have followed chuck hubers lasso tutorial steps and here is my command: lasso linear continuousdepvar $ifactors $vlcontinuous
but for some reason i get the error message: "depvar can not be specified in (alwaysvars) othervars"
could anyone tell me what i am doing wrong?
thanks
Mathieu

[Rich Goldstein]

"alwaysvars" and "othervars" each refer to predictors and thus the dependent variable (called "depvar" in the Stata message) should not be included; since you don't show us the definitions of the globals (please read the FAQ and follow it's advice), more specific information cannot be given

[William Lisowski]

To Rich's advice I will add that you should tell us where to find "chuck hubers lasso tutorial steps" that you are trying to follow.

[mathieu nacher]

hello, here is the link to the stata tutorial for lasso by chuck huber on stata 16, which i also use. i actually tried with the fake survey dataset he uses in the tutorial, i replicated the same commands he typed and i still get the same error message "depvar can not be specified in (alwaysvars) othervars"
https://www.youtube.com/watch?v=efYBzFcKWn8

[William Lisowski]

Here's the problem with the tutorial code copied from the tutorial video that you linked to.

When we webuse the example data

. webuse fakesurvey, clear
(Fictitious survey data)

. vl set, categorical(4) uncertain(0)

-------------------------------------------------------------------------------
                  |                      Macro's contents
                  |------------------------------------------------------------
Macro             |  # Vars   Description
------------------+------------------------------------------------------------
System            |
  $vlcategorical  |     115   categorical variables
  $vlcontinuous   |      47   continuous variables
  $vluncertain    |       0   perhaps continuous, perhaps categorical variables
  $vlother        |       9   all missing or constant variables
-------------------------------------------------------------------------------

we create the variable lists shown. However $vlcontinuous includes the variable q104 which will be used as the dependent variable in the lasso command, and we get the error message you saw.

. lasso linear q104 ($idemographics) $ifactors $vlcontinuous if sample==1, rseed(1234)
depvar may not be specified in (alwaysvars) othervars
r(198);

(And as an aside, we note that $idemographics was not defined in the video, but this is not in and of itself a problem, although it may be why my results do not match those in the video.)

The solution to the problem is to add after the vl set command

. vl drop (q104), system
note: 1 variable specified; 1 variable removed from vl system macros.

to drop the q104 variable from $vlcontinuous. Then the error message is eliminated and lasso runs successfully.

So you will need to add

vl drop (continuousdepvar), system

to your code for your model.

[mathieu nacher]

wonderful, thanks a lot!

[mathieu nacher]

sorry but i get this: vl drop(depvar), system
"drop(depvar) invalid vl subcommand"

[William Lisowski]

You have not proofread what you typed and compared it to the code presented in post #5; had you done so you would have seen that you omitted the space following the drop subcommand and and preceding the left parenthesis surrounding the variable list..

help vl

[mathieu nacher]

ok it works
thanks a lot