Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Yue Maggie Zhou <ymzhou@rhsmith.umd.edu> |
To | statalist <statalist@hsphsun2.harvard.edu> |
Subject | st: manual 2sls, xtivreg and xtivreg2 |
Date | Wed, 30 Jun 2010 10:58:18 -0400 |
Hi,
I have been puzzled by the different results I got when using manual 2sls, xtivreg and xtivreg2 on the same dataset with the same variables. It seems that manual 2sls, xtivreg and xtivreg2 report similar point estimates of the coefficients. However, there are differences in reported N (number of observation) and SE (standard errors). In particular, xtivreg2 reports fewer number of observations (N 5% smaller) though the estimated coefficients are similar. I guess the observations dropped are those with only one-year of data. In addition, manual2sls (with or without robust SE) reports the smallest SE, even smaller than xtivreg2 without robust SE.
Is this common or unique to my dataset?
Thanks,
Maggie