[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: 'Rolling' Slope Dummy

 From Robert Gaskell To statalist@hsphsun2.harvard.edu Subject Re: st: 'Rolling' Slope Dummy Date Sun, 6 Mar 2005 00:32:07 +0000

Thanks very much for that - your notes have been extremely helpful and
answered my question fully!  However, I have found that when
introducing these 'interaction effects between continuous variables',
if I include the slope term *or* the intercept term the coefficients
are significant, but when including both (as you suggest), the
coefficients are insignificant.  Does the intercept term need to be
included where the mid-point of the variable that is being added in is
zero, as in this case?

Best regards,
Robert Gaskell

On Fri, 04 Mar 2005 13:16:21 -0500, Richard. Williams
<Richard.A.Williams.5@nd.edu> wrote:
> At 12:20 PM 3/4/2005, Robert Gaskell wrote:
> >All,
> >
> >I am looking at the relationship between two variables, and am trying
> >to determine the effect that a third variable has on the coefficient
> >of this relationship.  Is it possible to use a sort of 'rolling' dummy
> >(this is I am sure a completely inaccurate name), whereby instead of
> >having a standard dummy:
> >
> >gen dummy = [some expression]
> >gen var2dummy = var2*dummy
> >reg var1 var2 var2dummy
>
> When working with interaction terms, you generally want to be including the
> lower-order terms used to compute the interactions.  e.g.
>
> reg var1 var2 dummy var2dummy
>
> Otherwise you'll find that arbitrary changes in the zero point of var2
> change your results (e.g. you'll get different results if your scale ranges
> from -3 to 3 than if it runs from 1 to 7). To see this, subtract a constant
> from var2 first and then re-run your above commands.  To put it another
> way, if you are going to let the slopes differ across groups, you generally
> want to let the intercepts differ too.
>
> Discussions of using interaction effects for group comparisons can be found at
>
> http://www.nd.edu/~rwilliam/stats2/l51.pdf
> http://www.nd.edu/~rwilliam/stats2/l52.pdf
>
> >...which only has a true/false which isn't suitable here, I multiply
> >var2 and var3 to create some sort of pseudo-rolling-slope-dummy:
> >
> >gen var2var3 = var2*var3
> >reg var1 var2 var2var3
> >
> >Here, I would take the interpritation to be that for every unit
> >increase in var3, the coefficient on var2 increases by the coefficient
> >on var2var3.
>
> Again, you probably want
>
> reg var1 var2 var3 var2var3
>
> If I understand you correctly, you are talking about interactions involving
> two continuous variables.  These can get a little confusing to
> interpret.  My own attempt to make sense of them can be found at
>
> http://www.nd.edu/~rwilliam/stats2/l55.pdf
>
> The sources cited (or any source that discusses interaction terms at
> length) may be worth reading if you need a bit more depth than these brief
> notes provide.
>
>
> Richard Williams, Associate Professor
> OFFICE: (574)631-6668, (574)631-6463
> FAX:    (574)288-4373
> HOME:   (574)289-5227
> EMAIL:  Richard.A.Williams.5@ND.Edu
> WWW (personal):    http://www.nd.edu/~rwilliam
> WWW (department):    http://www.nd.edu/~soc
>
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

 © Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index