Reduction in number of observation after regression model

[lao wu]

Dear Statalist members,

I face a problem regarding reduction in number of observation after using regression model. Actually i have more than 20,000 firm observations for almost all variables used in my regression model. I used Order Probit model. but after using that model results show number of observation as 15,667. I don't know why this is happening. Kindly suggest me any solution and cause for this.

I am very thankful.

[Nick Cox]

The number of observations is exactly what it was; the only question is which were used in your model. The most obvious reason will be missing values; missing values in one variable used are quite sufficient to exclude observations from a model fit.

edit if !e(sample)

after your model fit shows which were excluded.

[Carlo Lazzaro]

Iao:
have you got missing values in your dataset?
If that were the case, please note that Stata applies listwise deletion to all the the observations with missingness in any of the variables.

PS: crossed in the cyberspace with Nick's similar but more complete reply.

[Phil Bromiley]

Iao - You may find the misstable function helpful. It is designed to help you understand what is going on with missing data.

[Nick Cox]

Good advice, but misstable is a command, not a function. Functions and commands are distinct in Stata.