Help with Simple slopes

[Gee Gii]

Hi !
I’m hoping someone can help me! I’m running a regression with an interaction with 2 continuous predictor variables Ps and pc one continuous dv, flow and one covariant level of exp with 3 cats. I want to compare 4 subtypes made up of high/low Ps and high/low pc and there effect on flow. High would be +1sd and low would be -1 Sd. I think I need to create simple slopes for each subtype and then run contrasts between them! Does anyone know how to do this in stata! Any help is much appreciated!

[Erik Ruzek]

Can you show us the regression model in Stata? Please be sure to include the i. prefix in front of categorical and c. prefix in front of continuous variables?

In general, Stata's margins command is great for this. I'd suggest that you read the margins PDF documentation for more examples than you can possibly imagine.

[Bruce Weaver]

As Erik Ruzek notes, -margins- (with dydx) and -marginsplot- will be helpful. You can relevant examples here:

• https://stats.oarc.ucla.edu/stata/se...actions-stata/

Cheers,
Bruce

[Gee Gii]

Thanks so much for the reply! I ran this margins command and basically, I want to know if it is possible to compare these 4 groups, specifically compare 3 v 1, 3 v 2, 3 v 4.

. margins, at(ecp=(33.53841 52.40084) psp=(20.555251 29.847189)) vsquish
Predictive margins Number of obs = 164
Model VCE: OLS
Expression: Linear prediction, predict()
1._at: psp = 20.55525
ecp = 33.53841
2._at: psp = 20.55525
ecp = 52.40084
3._at: psp = 29.84719
ecp = 33.53841
4._at: psp = 29.84719
ecp = 52.40084
------------------------------------------------------------------------------
| Delta-method
| Margin std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
_at |
1 | 123.1813 1.838294 67.01 0.000 119.5505 126.8121
2 | 105.2 4.06512 25.88 0.000 97.17097 113.2289
3 | 131.3891 3.363699 39.06 0.000 124.7455 138.0328
4 | 130.2848 1.853819 70.28 0.000 126.6234 133.9463

[Gee Gii]

and this is the regression output.

regress flowtotal i.levelofexp c.pspcen##c.ecpcen, baselevels
Source | SS df MS Number of obs = 164
Model | 17121.4756 5 3424.29511 Prob > F = 0.0000
Residual | 37511.2744 158 237.413129 R-squared = 0.3134
-------------+---------------------------------- Adj R-squared = 0.2917
Total | 54632.75 163 335.170245 Root MSE = 15.408
---------------------------------------------------------------------------------------------
flowtotal | Coefficient Std. err. t P>|t| [95% conf. interval]
----------------------------+----------------------------------------------------------------
levelofexp |
0 | 0 (base)
1 | 11.28272 2.902751 3.89 0.000 5.54952 17.01592
2 | 16.1383 3.006963 5.37 0.000 10.19927 22.07732
|
pspcen | 1.791473 .3564857 5.03 0.000 1.08738 2.495565
ecpcen | -.5059386 .1745452 -2.90 0.004 -.8506814 -.1611958
|
c.pspcen#c.ecpce | .0962927 .0254272 3.79 0.000 .0460717 .1465137
|
_cons | 114.075 2.118144 53.86 0.000 109.8915 118.2586
---------------------------------------------------------------------------------------------

[Bruce Weaver]

Hello Gee Gii. Please post your output again, but this time, highlight it and then click on the # key in the menu bar. This will enclose it in code (i.e., text) delimiters making if much easier to read. Thanks.

[Erik Ruzek]

Here is an example. The key is that you want to use the contrast option in margins. I suggest reading the PDF documentation on margins, contrast.

use https://www.stata-press.com/data/r16/nlsw88


poisson wage i.union c.hours##c.ttl_exp, vce(robust)

note: you are responsible for interpretation of noncount dep. variable

Iteration 0:   log pseudolikelihood = -5086.0385  
Iteration 1:   log pseudolikelihood = -5086.0382  

Poisson regression                              Number of obs     =      1,877
                                                Wald chi2(4)      =     456.39
                                                Prob > chi2       =     0.0000
Log pseudolikelihood = -5086.0382               Pseudo R2         =     0.0617

-----------------------------------------------------------------------------------
                  |               Robust
             wage |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
            union |
           union  |   .1574339   .0262123     6.01   0.000     .1060587    .2088091
            hours |   .0045291   .0033695     1.34   0.179     -.002075    .0111331
          ttl_exp |   .0477448   .0110073     4.34   0.000      .026171    .0693187
                  |
c.hours#c.ttl_exp |  -.0001319   .0002807    -0.47   0.638    -.0006821    .0004183
                  |
            _cons |   1.243886   .1257394     9.89   0.000     .9974411    1.490331
-----------------------------------------------------------------------------------

Now the margins call, after first getting the mean and standard deviation of the variables of interest.

tabstat hours ttl_exp, s(mean sd) c(s)

    variable |      mean        sd
-------------+--------------------
       hours |  37.21811  10.50914
     ttl_exp |  12.53498  4.610208
----------------------------------

margins, at(hours = (26.71 47.73) ttl_exp = (7.93 17.13)) contrast(atcontrast(ib3.r._at)) 
Contrasts of predictive margins                 Number of obs     =      1,877
Model VCE    : Robust

Expression   : Predicted number of events, predict()

1._at        : hours           =       26.71
               ttl_exp         =        7.93

2._at        : hours           =       26.71
               ttl_exp         =       17.13

3._at        : hours           =       47.73
               ttl_exp         =        7.93

4._at        : hours           =       47.73
               ttl_exp         =       17.13

------------------------------------------------
             |         df        chi2     P>chi2
-------------+----------------------------------
         _at |
   (1 vs 3)  |          1        3.98     0.0460
   (2 vs 3)  |          1       45.47     0.0000
   (4 vs 3)  |          1      129.34     0.0000
      Joint  |          3      350.48     0.0000
------------------------------------------------

--------------------------------------------------------------
             |            Delta-method
             |   Contrast   Std. Err.     [95% Conf. Interval]
-------------+------------------------------------------------
         _at |
   (1 vs 3)  |  -.4400131   .2205108     -.8722064   -.0078198
   (2 vs 3)  |     2.4682   .3660223       1.75081    3.185591
   (4 vs 3)  |    2.89337   .2544092      2.394737    3.392003
--------------------------------------------------------------

[Gee Gii]

Ahhh amazing! Legends! this is perfect! I cannot thank you two enough. Erik Ruzek Bruce Weaver