Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: Plotting interactions
From
Amal Khanolkar <[email protected]>
To
"[email protected]" <[email protected]>
Subject
RE: st: Plotting interactions
Date
Mon, 30 Sep 2013 12:05:23 +0000
Hi David & John,
Yes - I agree the main effects should be included and the double '##' is an easy option to get this done.
Thanks for the tip Kostas - will look at those do files.
Thanks,
/Amal.
Amal Khanolkar, PhD candidate,
Centre for Health Equity Studies (CHESS),
Karolinska Institutet,
106 91 Stockholm.
Ph# +46(0)8 162584/+46(0)73 0899409
www.chess.su.se
________________________________________
From: [email protected] [[email protected]] on behalf of John Antonakis [[email protected]]
Sent: 30 September 2013 13:43
To: [email protected]
Subject: Re: st: Plotting interactions
Right......not including the main effects is tantamount to having an
omitted variable--the interaction will certainly correlate with its
constituents.
See:
Evans, M. G. 1991. The problem of analyzing multiplicative composites.
American Psychologist, 46(1): 6-15.
Best,
J.
__________________________________________
John Antonakis
Professor of Organizational Behavior
Director, Ph.D. Program in Management
Faculty of Business and Economics
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis
Associate Editor:
The Leadership Quarterly
Organizational Research Methods
__________________________________________
On 30.09.2013 13:34, David Hoaglin wrote:
> Hi, Amal.
>
> I won't speak to the subsequent programming, but it is unusual for the
> predictors in a regression model to include the interaction of two
> variables and not include the "main effect" of either of those
> variables. Would the results make better sense if your model included
> the main effects of ethnicity_bi2 and smoke2? You can include those
> and the two-variable interaction by using the ## operator instead of
> #. If you simply want to combine those two variables in a 6-category
> predictor (and have the first category, ethnicity_bi2 = 1 and smoke2 =
> 2, as part of the constant term), then that is what your current model
> does.
>
> David Hoaglin
>
> On Mon, Sep 30, 2013 at 4:41 AM, Amal Khanolkar <[email protected]> wrote:
>> Hi All,
>>
>> I'm trying to plot interactions post regression using the following syntax:
>>
>> The model:
>>
>> . regress bwtgestage_sd i.ethnicity_bi2#i.smoke2 sex ib2.magecat i.parity ib2.education i.famsit_new ib2.MBMI5 gestage_wk if multibirth==1, vce(robust)
>>
>> Linear regression Number of obs = 1144571
>> F( 22,1144548) = 4543.46
>> Prob > F = 0.0000
>> R-squared = 0.0820
>> Root MSE = .94207
>>
>> --------------------------------------------------------------------------------------
>> | Robust
>> bwtgestage_sd | Coef. Std. Err. t P>|t| [95% Conf. Interval]
>> ---------------------+----------------------------------------------------------------
>> ethnicity_bi2#smoke2 |
>> 1 3 | -.3777867 .0023854 -158.37 0.000 -.3824621 -.3731114
>> 2 2 | -.1160349 .0063877 -18.17 0.000 -.1285545 -.1035153
>> 2 3 | -.4195231 .0107743 -38.94 0.000 -.4406403 -.3984059
>> 3 2 | -.4354954 .0044767 -97.28 0.000 -.4442696 -.4267213
>> 3 3 | -.577776 .0153539 -37.63 0.000 -.6078691 -.547683
>> |
>> sex | .0127302 .0017618 7.23 0.000 .0092772 .0161832
>> |
>> magecat |
>> 1 | .0813647 .0062532 13.01 0.000 .0691086 .0936207
>> 3 | -.0483047 .0024836 -19.45 0.000 -.0531725 -.0434369
>> 4 | -.0756843 .0028012 -27.02 0.000 -.0811745 -.070194
>> 5 | -.1036756 .0037371 -27.74 0.000 -.1110003 -.096351
>> 6 | -.1385867 .0075963 -18.24 0.000 -.1534752 -.1236982
>> |
>> parity |
>> 2 | .308966 .0020534 150.47 0.000 .3049415 .3129905
>> 3 | .4244317 .0026556 159.83 0.000 .4192269 .4296365
>> |
>> education |
>> 1 | -.044489 .0029772 -14.94 0.000 -.0503241 -.0386539
>> 3 | .0334391 .0024809 13.48 0.000 .0285766 .0383016
>> 4 | .0482377 .0025474 18.94 0.000 .0432449 .0532304
>> |
>> famsit_new |
>> 3 | -.0321022 .0055319 -5.80 0.000 -.0429445 -.0212598
>> 4 | -.023603 .0077085 -3.06 0.002 -.0387113 -.0084947
>> |
>> MBMI5 |
>> 1 | -.2915349 .0039634 -73.56 0.000 -.2993029 -.2837669
>> 3 | .249358 .0023809 104.73 0.000 .2446914 .2540245
>> 4 | .3691747 .0042171 87.54 0.000 .3609092 .3774401
>> |
>> gestage_wktemp | -.0000848 .0005237 -0.16 0.871 -.0011111 .0009416
>> _cons | -.0348456 .0209644 -1.66 0.096 -.0759352 .006244
>>
>>
>> I then use the following :
>>
>> qui foreach x of var magecat {
>> sum `x', d
>> replace `x' = r(p50)
>> }
>> predict p
>> predict se, stdp
>> tw (line p ethnicity_bi2 if smoke2==2, sort) (line p ethnicity_bi2 if smoke2==3, sort)
>>
>>
>> I would like to know if the line above 'qui foreach x of var magecat' actually does indicate all categories of all variables magecat onwards as specified in the model including the continuous variable gestaga_wk?
>>
>> I don't seem to get a graph I was expecting - or at least I can't make sense of it. Have I specified the graph correctly or is there a better way plot interactions from a regression model?
>>
>> Thanks!
>> Amal
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/