Creating an S-curve graph based on regression results.

[Boryana Dimitrova]

Hi All,

I am trying to figure out how to graph an S-curve relationship between the DV (lnbannersales_store) and one of the IVs (countries_regions) in Stata. This is what my regression model looks like:

xtscc lnbannersales_store sfd lncheckouts countries_regions sqcountries_regions cubecountries_regions foreignmarketgrowth intspeed lnforeignsales globecultdist lnpoldist,fe

I would also like to show how the S-curve relationship changes once the relationship between tDV (lnbannersales_store) and one of the IVs (countries_regions) is moderated by sfd (sfd_countriesregions sfd_sqcontriesrregions sfd_cubecountriesregions).

xtscc lnbannersales_store sfd lncheckouts countries_regions sqcountries_regions cubecountries_regions foreignmarketgrowth intspeed lnforeignsales globecultdist lnpoldist sfd_countriesregions sfd_sqcontriesrregions sfd_cubecountriesregions,fe


Any ideas how I could create the graphs above would be greatly appreciated.


Thank you in advance!

Boryana

[Andrew Musau]

The S-curve relationship should be apparent from looking at your coefficients and their significance. You can always do a twoway plot as a first step, for example:

. webuse nhanes2

. desc hdresult

              storage   display    value
variable name   type    format     label      variable label
----------------------------------------------------------------------------------------
hdresult        int     %9.0g                 high density lipids (mg/dL)

. gen age3= age^3

. regress hdresult c.age##c.age age3

      Source |       SS           df       MS      Number of obs   =     8,720
-------------+----------------------------------   F(3, 8716)      =      7.47
       Model |  4577.28505         3  1525.76168   Prob > F        =    0.0001
    Residual |  1781305.25     8,716  204.371873   R-squared       =    0.0026
-------------+----------------------------------   Adj R-squared   =    0.0022
       Total |  1785882.53     8,719  204.826532   Root MSE        =    14.296

------------------------------------------------------------------------------
    hdresult |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |  -1.126969   .2978212    -3.78   0.000    -1.710769   -.5431695
             |
 c.age#c.age |   .0265066   .0067507     3.93   0.000     .0132736    .0397395
             |
        age3 |  -.0001893   .0000481    -3.94   0.000    -.0002835   -.0000951
       _cons |    63.6871   4.053082    15.71   0.000      55.7421     71.6321
------------------------------------------------------------------------------

. twoway lowess hdresult age

You can also plot predicted outcome against the regressor of interest.

predict yhat, xb
twoway lowess yhat age

[Boryana Dimitrova]

Thank you!