probit missing obs

[karamat khan]

hi all, im sory if my question is too obvious, here im using probit regression on cross sectional data like

probit y x, vce(robust)

in "x" except age size and education all are dummy variables which includes sector(2 dummies) , training(1 dummy) , export (1 dummy) size(3 dummy) etc
my observation numbers are 2676 and for different variables i have different obs no from 287 to 2676 (this extreme low obs are only for one variable) which means that som obs are missing for these variables

how stata output looks like?


on top of output it shows total OBSERVATIONS not more than 200 when i have enough observations for all variables , i dont understand why?? is it becz of missing values ? how to deal with this?

secondly, sector dummies are omitted from the result ? why they get omitted from the output ?

third, all other variables are insignificant except two of them, pseudo rsqr 9.5%, Lr chi 19.94..................... is it correct to go on with these results?

[Richard Williams]

Karamat, this might be easier to answer if you showed your commands and output using code tags. See pt 12 of the FAQ on asking questions effectively.

Remember, cases are dropped on a listwise basis -- if ANY dependent or independent variable has missing data, the entire case gets dropped. It sounds like at least some variables have a lot of missing data so that drives your overall N down.

[Phil Bromiley]

To extend Richard's comment, with one variable with only 287 observations, you're starting with a maximum of 287 observations, but if any of the other variables is missing for any of those 287 observations, you lose those observations as well. Clearly you have about 90 such missing values.