Test total effect after regression with interaction term.

[Gabriel Baldanzi]

Hello everyone,

After a mixed-effect model with an interaction term, I am trying to use -test- to compare the total effect (_b[exposure]+_b[exposure#covar]) vs the interaction-free effect. Also, I would like test if both coefficients are different than 0. However, I'm unsure if the output is what I was looking for. I would appreciate some help. I use Stata 14.2

More specifically, I have a biomarker measured at 3 age-points and exposure to NO2. I would like to determine with exposure to NO2 affects the biomarker level overall (repeated measure) and at the 3 age-points (not continuous). ID is the unique number for every individuals. The data is in long format
code:

 
.mixed biomarker c.NO2##i.age || ID: 
(output omitted)

.test _b[NO2] = _b[NO2] + _b[1.age#c.NO2] = _b[NO2] + _b[2.age#c.NO2] = 0   

 ( 1)  - [biomarker]1.age#c.NO2 = 0
 ( 2)  - [biomarker]2.age#c.NO2 = 0
 ( 3)  [biomarker]NO2 = 0

           chi2(  3) =   13.29
         Prob > chi2 =    0.0040

Is this -test- indeed considering the total effect? I'm unsure if the output:
" ( 1) - [biomarker]1.age#c.NO2 = 0"
really means what I intended:
"biomarker]NO2 + [biomarker]1.age#c.NO2 = 0"

Am I not writing the -test- correctly?

[Phil Bromiley]

Welcome to Stata list. You will increase your chances of useful answer by following the FAQ on asking questions – divide Stata code in code delimiters, readable Stata output, and sample data using dataex.

Normally when we talk about total effects we are talking about mediation not interactions.

_b[NO2] is the influence for age equals zero. _b[NO2] + _b[1.age#c.NO2] is the influence of N02 for age equals one._b[NO2] + _b[2.age#c.NO2] is the influence of N02 for age equals two.

The coefficients on the interactions test whether the influence differs from the age=0 group.

As I noted above, total effect is not really a meaningful concept far as I know in interaction terms.