How to solve this problem? "time variable must contain only integer values r(451)"

[Sumaira Ashraf]

Hi,

I am trying to run panel logistic regression. It is a time series data of firms.

The tsset results are:

. tsset countfirm year
panel variable: countfirm (unbalanced)
time variable: year, 2001 to 2015
delta: 1 unit

After this, i declare it as a time series with these commands
.tis year
iis countfirm


When i try to run fe model, I am getting this error.

. xtset status nita,fe
option fe not allowed
r(198);

When i try without fe, I am getting this error.

. xtset status nita
time variable must contain only integer values
r(451);


status is a dummy variable 0 and 1. I have a list of independent variables but i am trying to run with one variable first.
My data do not have duplicate years and do not have any jumps in time periods. Can anyone please help? I am badly stuck.

Regards,
Sumaira

[Clyde Schechter]

Are you perhaps confusing -xtset- with -xtreg-? You have already -tsset- your data with a panel and time variables, the results of which are entirely equivalent to having -xtset- the data with the same panel and time variables. You then have a new -xtset status nita- command and you indicate that status is a 0/1 variable. But it is very odd indeed to try to define panels with a single 0/1 variable: usually there are many more than 2 panels in a data set. And you don't say at all waht the variable nita is, but if it doesn't contain integer values, it is not suitable to be used as the time variable when -xtset-ing (nor -tsset-ing the data. The fact that you erroneously initially tried this with an -fe- option makes me think you were trying to do a fixed-effects regression. But -xtset- doesn't do that: you mean -xtreg- if that's what you want.

[William Lisowski]

Yes, you need variation in the dependent variable for any regression in order to determine the effects of the independent variables. Otherwise, all you know is that none of your independent variables have any effect on the dependent variable. If you omit the fixed effects, then you will have variation across the firms, or countries, or whatever it is your panels are.

[Richard Williams]

You can’t use fe if there is no variation within groups. For a brief explanation, see https://www3.nd.edu/~rwilliam/xsoc73...xedEffects.pdf

[Marcos Almeida]

status is a dummy variable 0 and 1. I have a list of independent variables but i am trying to run with one variable first.

nita is a financial ratio "Net income/total assets" I have a set of 17 more ratios

Little wonder you got an error messag when you - xtset status nita - for the first time.

The correct command as you can check in the Stata Manual is something like:

. xtset panelvar timevar

If I understood your case correctly, neither "status" is a panel variable nor "nita" is a time variable.