nlogit

[Ghia Osseiran]

Hi all- I am trying to run a nested logit model where I'm looking at the determinants of graduate employment and wanted to use the nested logit to account for self-selection bias in the selection of fields of study.
I have fields of study at the bottom of the hierarchy (with each field as a dummy variable) and have created the variable grad_id= id if grad==1

If I run the below command:
nlogit employed X (X=vector of independent variables other than field of study)
|| type: fieldofstudy, base(grad) ||
||:, noconst case(grad_id)

I get the error:
option case() required

Would appreciate any tips on how to resolve this.

Thanks,
Ghia

[Rich Goldstein]

if I put your unclear syntax above into one line, I get:

nlogit employed X || type: fieldofstudy, base(grad) || ||:, noconst case(grad_id)

which appears wrong: you have two sets of "||" together and you have only a ":" following the last set; how about trying:

nlogit employed X || type: fieldofstudy, base(grad) , noconst case(grad_id)

note that I have made a guess re: where your typo is; maybe the typo is that you left something out between the contiguous pairs of "||" and/or you left out something prior to the last colon????

[Ghia Osseiran]

Thanks Rich now the error I get is: invalid 'noconst'
equation 2 defining level 1 is improperly specified

[Rich Goldstein]

what EXACTLY did you type - and please put it within CODE blocks (read the FAQ to find out how)

actually, it looks as though I had a typo in my second version (if that's what you typed) as I have 2 commas and there should only be one (before any options) (and, yes, I know that one can have more than one comma but options always follow odd-numbered commas and in my previous message I have options after comma number 2)

[Richard Williams]

actually, it looks as though I had a typo in my second version (if that's what you typed) as I have 2 commas and there should only be one (before any options) (and, yes, I know that one can have more than one comma but options always follow odd-numbered commas and in my previous message I have options after comma number 2)

As a sidelight, it annoys me that Stata whines about the use of a 2nd comma. It ought to just ignore it and assume this is yet another option.

[Ghia Osseiran]

for fieldsofstudy, where there are a total of 8 fields, I have the code:

nlogitgen type= fieldsofstudy (100, 200, 300, 450, 500, 600, 700, 800)


For the logit I have:

nlogit employed $X || type: hatfield1, base(grad) , noconst case(grad_id)

$X being a macro for a series of independent variables
grad is a dummy variable for graduate
grad_id=id if grad==1

And this is the response I'm getting:
invalid 'noconst'
equation 2 defining level 1 is improperly specified
r(198);

[Rich Goldstein]

first to Ghia: please use CODE blocks (read the FAQ); also, note that, as far as I can see, you still have an extra comma; delete that second comma as I previously said

second to Richard: actually, it is sometimes handy (as I play with a command, to have the ability to use more than one comma; e.g., there is nothing wrong with:

sysuse auto
ttest price, by(foreign), if rep78<., une

[Ghia Osseiran]

Thank you for your feedback. Essentially I have three choice levels, the first is graduate self-selection into field of study, the second is graduate employment versus unemployment, and the third is graduate overqualification in the labour market contingent on employment. I first want to use nested logit in order to account for self-selection into field of study. If I run the below:

nlogitgen type= fieldofstudy (100, 200, 300, 450, 500, 600, 700, 800)

nlogit employed $X|| type: fieldofstudy, base(grad) noconst case(grad_id)

I get:

note: nonnested model, results are the same as clogit
note: 23053 cases dropped because they have only one alternative
note: 1832 cases (3750 obs) dropped due to no positive outcome or multiple positive
outcomes per case
no cases remain after removing invalid observations

Would appreciate if anyone can spot light on the error here.

Thanks,
Ghia

[Meriem OUD]

Hey Ghia

I have the same problem did you find the solution

[Embarika Farouk]

I have the same problem, anyone got the solution

[Alina Faruk]

Hi anyone found the solution? Facing the same problem.