Estimating CRRA Utility Function - Getting "variable __000000 not found"

[Ash Belur]

I am a rookie working with STATA, but am trying to estimate a CRRA Exponential Utility function using some sample test dataset.
I am getting the error "variable __000000 not found"

Can anyone help?


clear

input Choices P0left P1left P0right P1right prize0l prize1l prize0r prize1r
1 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
0 1 0 0.50 0.50 1500 1500 1500 9000
0 1 0 0.50 0.50 1500 1500 1500 9000
0 1 0 0.50 0.50 1500 1500 1500 9000
0 1 0 0.50 0.50 1500 1500 1500 9000
0 1 0 0.50 0.50 1500 1500 1500 9000
1 1 0 0.50 0.50 1500 1500 1500 9000
end

capture program drop ML_eut0
program define ML_eut0

args lnf r

tempvar prob0l prob1l prob0r prob1r y0l y1l y0r y1r
tempvar euL euR euDiff
tempvar lnf euRatio lnf_eut lnf_pt p1 p2 f1 f2

quietly {
generate double `prob0l' = $ML_y2
generate double `prob1l' = $ML_y3

generate double `prob0r' = $ML_y4
generate double `prob1r' = $ML_y5

generate double `y0l' = ( $ML_y6 ) ^`r'
generate double `y1l' = ( $ML_y7 ) ^`r'
generate double `y0r' = ( $ML_y8 ) ^`r'
generate double `y1r' = ( $ML_y9 ) ^`r'

// generate `euL' = (`prob0l' * `y0l') + (`prob1l' * `y1l')
// generate `euR' = (`prob0r' * `y0r') + (`prob1r' * `y1r')
generate `euL' = (1 * `y0l') + (0 * `y1l')
generate `euR' = (0.5 * `y0r') + (0.5 * `y1r')
generate double `euDiff' = `euR' - `euL'

replace `lnf' = ln(normal( `euDiff')) if $ML_y1==1
replace `lnf' = ln(normal(-`euDiff')) if $ML_y1==0
}
end

ml model lf ML_eut0 (r: Choices P0left P1left P0right P1right prize0l prize1l prize0r prize1r = ) if Choices~=., technique(nr) maximize

[Nick Cox]

Variable names such as __000000 starting with double underscore and followed by integers 000000, 000001, ... generally denote temporary variables.

So, two questions:

1. Exactly where does the error message arise, that is, after which command? You may need to

set trace on

to see that.


2. How are you defining your program? In essence, the code needs to be run all at once, not in chunks.

[Ash Belur]

Nick Cox - The first question / hint was very helpful. It indicated an error related to the second "replace lnf" command.
This led me to find that my tempvar incorrectly included lnf as a temporary variable - which it should not have.

Thanks - let's see how far I get on the real data.
It did run all the way through and converged.