Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | "Watson, Kimberly" <kawatson@rti.org> |
| To | <statalist@hsphsun2.harvard.edu> |
| Subject | st: "Partial" Triple Interaction in Logit Model Not Supported by INTEFF3 |
| Date | Thu, 28 Jun 2012 13:32:07 -0400 |
I have two questions related to INTEFF3:
(1) Has a macro been written to accommodate a "partial" triple
interaction?
I say "partial" because I want to interact a 3-category variable and a
binary variable. INTEFF isn't appropriate because my first variable is
neither continuous or binary. Looking at the documentation for INTEFF3,
all interaction combinations must be present (I'm missing cat2*cat3 and
cat2*cat3*dummy because they're both equal to 0). My model is
generically "svy: logit y cat2 cat3 dummy cat2*dummy cat3*dummy;"
(2) Any suggestions for determining the source of a conformability
error?
Even though my situation doesn't quite work for INTEFF3, I wanted to
investigate it's use for future interaction questions. So, I created
meaningful dummies based on the documented syntax
(http://fmwww.bc.edu/RePEc/bocode/i/inteff3.html using x1, x2, and x3).
I do not have missing data and none of the variables or their
interactions are dropped due to collinearity. However, I keep getting a
conformability error r(503). I've tried resetting the version as
suggested in older posts. Also, I've tried running as logit in case
weights aren't supported.
Summary of Data:
+--------------------------------------------------------+
| y x1 x2 x3 x1x2 x1x3 x2x3 x1x2x3 count |
|--------------------------------------------------------|
| 0 0 0 0 0 0 0 0 964 |
| 0 0 0 1 0 0 0 0 554 |
| 0 0 1 0 0 0 0 0 966 |
| 0 0 1 1 0 0 1 0 545 |
| 0 1 0 0 0 0 0 0 1747 |
| 0 1 0 1 0 1 0 0 1149 |
| 0 1 1 0 1 0 0 0 2059 |
| 0 1 1 1 1 1 1 1 1287 |
|--------------------------------------------------------|
| 1 0 0 0 0 0 0 0 341 |
| 1 0 0 1 0 0 0 0 75 |
| 1 0 1 0 0 0 0 0 278 |
| 1 0 1 1 0 0 1 0 54 |
| 1 1 0 0 0 0 0 0 460 |
| 1 1 0 1 0 1 0 0 177 |
| 1 1 1 0 1 0 0 0 513 |
| 1 1 1 1 1 1 1 1 144 |
|--------------------------------------------------------|
Log Results Using Above Data:
. logit y x1 x2 x3 x1x2 x1x3 x2x3 x1x2x3;
Iteration 0: log likelihood = -5341.4475
Iteration 1: log likelihood = -5228.0162
Iteration 2: log likelihood = -5224.7923
Iteration 3: log likelihood = -5224.7842
Iteration 4: log likelihood = -5224.7842
Logistic regression Number of obs =
11313
LR chi2(7) =
233.33
Prob > chi2 =
0.0000
Log likelihood = -5224.7842 Pseudo R2 =
0.0218
------------------------------------------------------------------------
------
y | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------
------
x1 | -.29522 .0819525 -3.60 0.000 -.455844
-.134596
x2 | -.2063339 .0927481 -2.22 0.026 -.3881169
-.0245509
x3 | -.9604678 .1382327 -6.95 0.000 -1.231399
-.6895366
x1x2 | .1510629 .1174033 1.29 0.198 -.0790433
.381169
x1x3 | .424399 .1684478 2.52 0.012 .0942475
.7545506
x2x3 | -.1057913 .2099857 -0.50 0.614 -.5173556
.3057731
x1x2x3 | -.1586961 .2520254 -0.63 0.529 -.6526567
.3352646
_cons | -1.039209 .0630071 -16.49 0.000 -1.1627
-.9157172
------------------------------------------------------------------------
------
. inteff3;
conformability error
r(503);
Currently Running:
Stata/MP 12.1 for Unix (Linux 64-bit x86-64)
Revision 04 Apr 2012
Thanks!
K. Watson
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/