Fixed Effects Ordered Logit Regression

[olivia schuter]

Dear all,

I really need some help with regard to using the Blow up and Cluster estimator in order to estimate a fixed effects ordered logit regression.
To do so, I installed the program "feologit_buc"

Has anyone used this program on his data and could tell me how he did it?

My data is a panel data where country is the individual identifier and the observations are over years.

Thanks so much!
Kind regards.

[Richard Williams]

Where did you get this program? It doesn't show up with -findit-, so I assume you got it from somebody's web site?

[olivia schuter]

Hey there, you can find it in the appendix of the working paper consistent estimation of the fixed effects ordered models from the authors: Baetschmann, Staub and Winkelmann. I attached the pdf.

Some other people used the program with success but in none paper i find how they applied it to their own data.

Thanks in advance for you help.

[Richard Williams]

The paper (or a revised version of it) was published in

http://onlinelibrary.wiley.com/doi/1...ssa.12090/full

The programs and data are linked to at

http://onlinelibrary.wiley.com/journ...page/178_3.htm

I suggest going through these first and see if they address your problems, e.g. maybe the programs have been improved or it is clearer how to run them.

You can also find papers that cite the paper at

https://scholar.google.com/scholar?h...%2C15&as_sdtp=

I always thought fe oprobit couldn't be done. Maybe this paper proves otherwise, but you might see what others say about it.

[olivia schuter]

Okay thanks a lot! I will try to go through it and see whether it can help me.

By chance, do you know whether I have to adapt the "version command" from the program code to my version from Stata? I use stata 13....

Kind regards

[Richard Williams]

Assuming their version was earlier than Stata 13, you will most likely want to leave their version command as is. The whole point of version commands is to keep your program running in case later versions of Stata include changes that would break your program if you didn't have version control.

[olivia schuter]

all right! Thank you

[olivia schuter]

Hello again,
If I use the following program, it is now working

program feologit_buc, eclass
preserve
gettoken gid 0: 0
gettoken y x: 0
qui sum `y'
local lk= r(min)
local hk= r(max)
tempvar iid id cid gidcid dk
bys `gid': gen `iid'=_n
gen long `id'=(`gid'*100)+`iid'
qui expand `=`hk'-`lk''
bys `id': gen `cid'=_n
gen long `gidcid'= `gid'*100+`cid'
qui gen `dk'= `y'>=`cid'+1
clogit `dk' `x', group(`gidcid') cluster(`gid')
mat b=e(b)
mat V=e(V)
sca N_g=e(N_clust)
ereturn clear
ereturn post b V
estadd sca N_g=N_g
end
feologit_buc country A170 log_pcBIP

Everything is working fine exceot for my Stata does not recognize the estadd - command....Do I have to use an ado for this command?


And secondly: From another paper of Baetschmann I got an adaptation of the feologit_buc program which also estimates the thresholds of the ordinal variable: You can find the Stata command in the attached file. If I want to use this program, I still get error messages that my "y" has an invalid name. Do you know how to solve this problem?

Thanks in advance!

[Marcos Almeida]

The command estadd can be done after installing the user-written estout: http://repec.org/bocode/e/estout/estadd.html

[olivia schuter]

thanks

[Richard Williams]

It would help if you show us exactly what you are typing and how Stata is responding. It may be an error in the program, but it may also be an error in how you are trying to run the program. There may be an error in how you copied the program, so show it too. Use code tags (see pt 12 in the FAQ).

[olivia schuter]

I just found my error... I didnt use the correct signs as for example ' and `. But now its working! Thanks for your help!