Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Stata 13 Syntax for 3-way interaction two continous variables and one dichotomous variable

    Hello everyone,

    I am going through the Stata 13 syntax code for creating interaction effects (margins). Previously, I have used user-created code for estimating marginal effects and the accompanying standard errors taken from Franzese/Kam 2007 (Modeling and Interpreting Interactive Hypotheses in Regression Analysis). Since this procedure is quite cumbersome and time-intensive, I am trying to switch to the stata routines, especially since there are more readily available plots.

    What I want to obtain is the following: Estimate a TSCS model (xtpcse, corr(ar1)), estimate and plot 3 two-way interactions (one between two continous variables, two between one continous and one dichotomous variable and plot the marginal effect alongside a histogram, which displays the distribution of the cases in order to gauge the substantive effects) and finally estimate and plot a three-way interaction between two continous and one dichotomos (dummy) variable.

    Here is my code from an example code (simple regression model), which I use to illustrate my example:


    reg zup_tota mlg corporat c.mlg#c.corporat gdpchang

    summ mlg
    local min1 = r(min)
    local max1 = r(max)


    local interval1 = round( (`max1'-`min1')/100, 0.01)

    local range1 ""`min1'(`interval1')`max1'""

    local at1 ""mlg = ("`range1'")""

    margins, dydx(corporat) at ( `at1' ) vsquish

    matrix t=r(table)'
    capture: drop line* at*
    svmat t, names(line)
    egen at1 = fill(""`range1'"")
    replace at1 = . if line1==.


    graph twoway (histogram mlg, yaxis(2) color(gs12) ) (line line1 at1, yaxis(1) lpattern(solid) lwidth(medthick) )
    (rline line5 line6 at1, yaxis(1) lpattern(dash) ),

    graph export marginsplot_d_Success1_p1Gdgg_p1Gdvp.emf, replace

    This is the code for a two-way continous variable interaction alongside a plot with a histogram.

    How would I go about an interaction effect with a dichotomous (dummy) variable? I wouldn't need a plot of a marginal effect along a moderating variable, since the dummy variable only has two possible manifestations.

    I think a plot similar to the one in the appendix would be best, how can I implement that in Stata 13 in light of the code I already have above? If I have that, I can continue with the code on the three-way interaction.

    Any help is greatly appreciated!




  • #2
    See help marginsplot.

    Michael N. Mitchell (2012) has a fantastic book with tons of examples on this.

    Best
    Daniel


    Mitchell, Michael N. (2012). Interpreting and Visualizing Regression Models Using Stata. College Station ,TX: Stata Press.
    (http://www.stata.com/bookstore/inter...ession-models/)

    Comment

    Working...
    X