 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: Post-estimation after MI

 From Tony Robertson To statalist@hsphsun2.harvard.edu Subject st: Post-estimation after MI Date Tue, 21 Aug 2012 11:55:30 +0100

```Hi

```
I am trying to compare two models, a saturated model (line 5) vs. simpler/theory-specific model model (line 4), with a simple Wald test following multiple imputation. I use the below syntax in the complete-case analysis, but am having difficulty when replicating following MI. I am using STATA 11.2. I am used to using the 'mim' command following MI with 'test' swapped for 'testparm'. However, applying the constraints in line 10 (x1=x2) is not allowed. I then tried to use mi estimate but again no success. Is there something else I could try or have missed/misunderstood?!
```
Thanks
Tony

complete-case syntax:

1. constraint 1 x1=x2
2. constraint 2 x2=x3
3. svy: cnsreg y a b x1 x2 x3 if miss==1, constraints(1 2)
4. svy: regress y a b x1 x2 x3 x12 x13 x23 x23 if miss==1
5. test (x12) (x13) (x23) (x123) (x1=x2) (x2=x3)

mim syntax:

6.   constraint 1 x1=x2
7.   constraint 2 x2=x3
8.   mim, cat(fit): svy: cnsreg y a b x1 x2 x3, constraints(1 2)
9.   mim: svy: regress y a b x1 x2 x3 x12 x13 x23 x23
10. mim: testparm (x12) (x13) (x23) (x123) (x1=x2) (x2=x3)

mi estimate syntax:

11. constraint 1 x1=x2
12. constraint 2 x2=x3
13. mi estimate: svy: cnsreg y a b x1 x2 x3, constraints(1 2)
14. mi estimate: svy: regress y a b x1 x2 x3 x12 x13 x23 x23
15. mi test (x12) (x13) (x23) (x123) (x1=x2) (x2=x3)

--
Tony Robertson
Career Development Fellow
MRC/CSO Social & Public Health Sciences Unit
4 Lilybank Gardens
Glasgow
G12 8RZ

Phone:   0141 357 7565