I'm a beginner when it comes to both statistics and STATA. This question might be more of a statistics question than a STATA question, but I'm looking for some help to interpret the results of this model and specifically the interaction of two continuous variables.
The following regression is from an assignment I'm working on. The dependent variable is wage with a plethora of explanatory variables. I know interactions can be added by using ##, but we were specifically asked to create the interactions the old-fashioned way and then add them to the model.
Code:
Source | SS df MS Number of obs = 550 -------------+---------------------------------- F(19, 530) = 22.82 Model | 59.3445613 19 3.12339796 Prob > F = 0.0000 Residual | 72.5548162 530 .13689588 R-squared = 0.4499 -------------+---------------------------------- Adj R-squared = 0.4302 Total | 131.899377 549 .240253875 Root MSE = .36999 -------------------------------------------------------------------------------------- ln_Wage | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------------+---------------------------------------------------------------- Education | .066528 .0143397 4.64 0.000 .0383583 .0946977 Experience | .0374092 .0111086 3.37 0.001 .015587 .0592314 Experience2 | -.0004465 .0001462 -3.05 0.002 -.0007337 -.0001593 Female | .0390759 .2135387 0.18 0.855 -.3804102 .4585621 Married | .0391859 .0523632 0.75 0.455 -.063679 .1420507 Union | .2220958 .0450569 4.93 0.000 .1335837 .3106079 Female_Education | -.0083708 .0144876 -0.58 0.564 -.0368309 .0200893 Female_Experience | -.0227384 .0102753 -2.21 0.027 -.0429237 -.0025531 Female_Experience2 | .0004476 .0002256 1.98 0.048 4.35e-06 .0008908 Female_Married | -.0479351 .0755239 -0.63 0.526 -.1962982 .1004279 Female_Union | .0834588 .0767746 1.09 0.278 -.0673611 .2342786 Nonwhite | -.1297967 .054883 -2.36 0.018 -.2376116 -.0219818 Hispanic | .0042826 .0675995 0.06 0.950 -.1285133 .1370785 Experience_Education | -.0003983 .0004914 -0.81 0.418 -.0013636 .000567 Manager | .2656318 .0625375 4.25 0.000 .14278 .3884837 Sales | .0954787 .078817 1.21 0.226 -.0593533 .2503108 Clerical | -.0133465 .050185 -0.27 0.790 -.1119324 .0852394 Service | -.2393058 .0618983 -3.87 0.000 -.360902 -.1177097 Professional | .118085 .0596035 1.98 0.048 .0009969 .2351731 _cons | .4570651 .1980329 2.31 0.021 .0680394 .8460907
I'm a little bit insecure and uncertain in how to interpret the results, but this is how I think it should be:
1. Take the variable Married for example. Holding all other variables constant, then being married (coded as 1) will increase wage by 3.9% compared to not being married.
2. Now, given then interaction Female_Married, it is also possible to say the effect of being married for females is 0.039 + (-0.0479351).
3. The big question I have is how Experience_Education affects the model (both are continuous). If we start by just looking at Education, then I would say that, again all other variables constant at 0, a one unit increase in Education will lead to an increase in wage by 0.066528 or 6.6%.
However, if we start to consider the interaction of Experience_Education, would the effect of Education then be, using the same approach as for the Female_Married interaction, 0.066528 + (-0.0003983)? This seems wrong to me and from what I found using Google it is not possible to interpret a continuous interaction this way, but other commands and plotting is necessary (something we have not covered in my course).
So, just from this model and without doing any other commands, what can possibly be said about Experience_Education? Because the way I'm thinking of it, when interpreting the coefficient of Education as 0.066528, then all other coefficients in the model are 0, including Experience_Education.
Would it be correct to say for each unit change in education, the slope of experience vs. wage decreases by -0.0003983? And the same effect on the slope of education vs. wage for each unit change in experience?
Any help or input is much appreciated. Thanks!
0 Response to Linear Regression: Interpreting Continuous Interaction?
Post a Comment