Announcement

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

  • Graphing Interactions for Visual Inspection

    Hello All,

    I am having difficulty fitting a interaction graph over a histogram.

    The primary issue is the x-axis extends beyond the observations that I have on the histogram. Additionally, the right y-axis should be the variable TE with proper magnitude.



    The code I am using i as follows (this is the complete code):
    xtgls flnfdiflow96dollars bitsumoecdweighted lnrgdpch lnpop econgrowth inflation resourcedepletion agreementwithdevcountry wtomemberdummy polcon3 bitsumoecdweightedpolcon3 te ie de c.lnrgdpch#c.te c.ie#c.lnrgdpch c.de#c.lnrgdpch, corr(ar1) force

    sort lnrgdpch
    matrix b=e(b)
    matrix V=e(V)

    scalar b2=b[1,2]
    scalar b14=b[1,14]

    scalar varb2=V[2,2]
    scalar varb14=V[14,14]

    scalar covb2b14=V[2,14]


    scalar list b2 b14 varb2 varb14 covb2b14

    generate MVZ=((_n-6)/100)

    replace MVZ=. if _n>2720

    gen conbx=b2+b14*MVZ if _n<2720

    gen consx=sqrt(varb2+varb14*(MVZ^2)+2*covb2b14*MVZ)
    gen ax=1.96*consx
    gen upperx=conbx+ax
    gen lowerx=conbx-ax

    gen where=-0.045 *****This is potentially where the problem lies****
    gen pipe = "|"
    egen tag_te = tag(te)
    egen tag_gdp = tag(lnrgdpch)

    gen yline=0

    graph twoway hist lnrgdpch, width(0.5) percent color(gs14) yaxis(2)
    || scatter where lnrgdpch if tag_gdp, plotr(m(b 4)) ms(none) mlabcolor(gs5) mlabel(pipe) mlabpos(6) legend(off)
    || line conbx MVZ, clpattern(solid) clwidth(medium) clcolor(black) yaxis(1),
    || line upperx MVZ, clpattern(dash) clwidth(thin) clcolor(black)
    || line lowerx MVZ, clpattern(dash) clwidth(thin) clcolor(black)
    || line yline MVZ, clwidth(thin) clcolor(black) clpattern(solid)
    || ,
    xlabel(-2 0 2 4 6 8 10 12, nogrid labsize(2))
    ylabel(-0.05 0 .05 .1 .15 .2 .25, axis(1) nogrid labsize(2))
    ylabel(0 2 4 6 8 10 12 14 16, axis(2) nogrid labsize(2))


    I apologize for the excessive code. But I realize my issues with the graph may stem from how I built the interaction and CIs.

    Any help would be great.

    Best,
    Jesse


  • #2
    Click image for larger version

Name:	Graph.jpg
Views:	1
Size:	24.3 KB
ID:	1293041
    Reposting graph

    Comment


    • #3
      The range of the x-axis can be limited by adding if conditions to the line commands. For example:
      Code:
      line conbx MVZ if MVZ>6 & MVZ<11
      There is no variable TE in your graph command.

      Comment


      • #4
        Hi Friedrich,

        thank you for your reply.

        You will notice that TE is in the interaction effect that is being graphed. When I run with your suggestion it simply shortens the line but does not condense the graph.

        Comment


        • #5
          Please show your code and the resulting graph (see Statalist FAQ, section 12).
          Last edited by Friedrich Huebler; 04 May 2015, 07:12.

          Comment

          Working...
          X