Hello everyone,
I am estimating a discontinuous growth curve model for a study about the impact of a treatment. All patients received the treatment, and had 3 pre measurements and 2 post measurements. Outcome is continuous. Model has a random intercept for the individual. The main variables in model: post (dummy for post period), time (time from start of study) and their interaction; these variables relate to treatment effect. I also have an additional demographic continuous variable, which is measured at the start of the study.
My question is actually about interpreting the demographic continuous variable called "cont". It is negative. Can I say for example: "cont is negatively associated with the outcome variable, such that patients with higher values of cont have lower values of the outcome? An increase of cont by one point leads to a 0.52 decrease in the outcome variable?"
I have shown the code below.
thanks!
I am estimating a discontinuous growth curve model for a study about the impact of a treatment. All patients received the treatment, and had 3 pre measurements and 2 post measurements. Outcome is continuous. Model has a random intercept for the individual. The main variables in model: post (dummy for post period), time (time from start of study) and their interaction; these variables relate to treatment effect. I also have an additional demographic continuous variable, which is measured at the start of the study.
My question is actually about interpreting the demographic continuous variable called "cont". It is negative. Can I say for example: "cont is negatively associated with the outcome variable, such that patients with higher values of cont have lower values of the outcome? An increase of cont by one point leads to a 0.52 decrease in the outcome variable?"
I have shown the code below.
Code:
. mixed outcome post_dummy time time_post cont || patient:
Performing EM optimization:
Performing gradient-based optimization:
Iteration 0: log likelihood = -509.03916
Iteration 1: log likelihood = -509.03916
Computing standard errors:
Mixed-effects ML regression Number of obs = 170
Group variable: patient Number of groups = 34
Obs per group:
min = 5
avg = 5.0
max = 5
Wald chi2(4) = 296.03
Log likelihood = -509.03916 Prob > chi2 = 0.0000
------------------------------------------------------------------------------
outcome | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
post_dummy | 15.17095 1.542634 9.83 0.000 12.14744 18.19446
time | .1499465 .5509066 0.27 0.785 -.9298107 1.229704
time_post | -.5301381 .5660319 -0.94 0.349 -1.63954 .579264
cont | -.52041 .2454857 -2.12 0.034 -1.001553 -.0392668
_cons | 85.29038 4.588404 18.59 0.000 76.29727 94.28348
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
patient: Identity |
var(_cons) | 6.663114 2.585102 3.114834 14.25344
-----------------------------+------------------------------------------------
var(Residual) | 19.0831 2.314476 15.04568 24.20395
------------------------------------------------------------------------------
LR test vs. linear model: chibar2(01) = 16.53 Prob >= chibar2 = 0.0000
.
. margins, at(cont = (13 17 21))
Predictive margins Number of obs = 170
Expression : Linear prediction, fixed portion, predict()
1._at : cont = 13
2._at : cont = 17
3._at : cont = 21
------------------------------------------------------------------------------
| Delta-method
| Margin Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_at |
1 | 83.23216 1.426538 58.35 0.000 80.43619 86.02812
2 | 81.15052 .6469439 125.44 0.000 79.88253 82.4185
3 | 79.06888 .8546855 92.51 0.000 77.39372 80.74403
------------------------------------------------------------------------------
