Aim 1a
Dan Leibel
3/11/2019
Significant 2-way interaction of CES x Age with HgbA1C, indicating that the CES predicts change in A1C over time.
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## HgbA1C ~ CES * cenAge + cenSex + cenRace + cenPovStat + cenEducnum +
## (cenAge | HNDid)
## Data: Tasneem_1a
## Control: lmerControl(check.nobs.vs.nRE = "ignore")
##
## REML criterion at convergence: 7778.3
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -5.3528 -0.3154 -0.0540 0.2228 9.5308
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## HNDid (Intercept) 0.2808797 0.52998
## cenAge 0.0002983 0.01727 1.00
## Residual 0.1884809 0.43414
## Number of obs: 4278, groups: HNDid, 1942
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 5.715e+00 2.367e-02 1.497e+03 241.436 < 2e-16 ***
## CES 1.617e-03 1.305e-03 1.502e+03 1.239 0.2154
## cenAge 1.329e-02 1.968e-03 3.228e+03 6.756 1.67e-11 ***
## cenSex -1.433e-02 2.631e-02 1.946e+03 -0.545 0.5860
## cenRace 2.287e-01 2.676e-02 1.964e+03 8.548 < 2e-16 ***
## cenPovStat -2.843e-02 2.811e-02 1.942e+03 -1.012 0.3119
## cenEducnum -9.527e-03 5.361e-03 1.957e+03 -1.777 0.0757 .
## CES:cenAge 2.501e-04 1.117e-04 3.403e+03 2.239 0.0252 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) CES cenAge cenSex cenRac cnPvSt cnEdcn
## CES -0.795
## cenAge 0.374 -0.311
## cenSex -0.043 0.089 -0.008
## cenRace -0.038 0.045 -0.005 -0.017
## cenPovStat 0.086 -0.109 0.036 0.044 -0.173
## cenEducnum -0.167 0.161 0.002 0.050 0.050 0.231
## CES:cenAge -0.297 0.419 -0.786 0.006 0.008 -0.007 0.011This effect later attenuated after adding the diabetes risk factors (not shown here), but I was able to figure out that this in part related to the lower n.
Plotting the interaction effect shows that greater depression is associated with the steepest increase in HgbA1C over time 
Next, we have a significant 2-way interaction of CES x Age with Glucose, indicating that the CES predicts change in glucose over time
## Linear mixed model fit by REML. t-tests use Satterthwaite's method [
## lmerModLmerTest]
## Formula:
## Glucose ~ CES * cenAge + cenSex + cenRace + cenPovStat + cenEducnum +
## (cenAge | HNDid)
## Data: Tasneem_1a
## Control: lmerControl(check.nobs.vs.nRE = "ignore")
##
## REML criterion at convergence: 36607.5
##
## Scaled residuals:
## Min 1Q Median 3Q Max
## -6.3511 -0.3394 -0.0626 0.2373 12.0240
##
## Random effects:
## Groups Name Variance Std.Dev. Corr
## HNDid (Intercept) 192.4389 13.8722
## cenAge 0.2827 0.5317 1.00
## Residual 172.5070 13.1342
## Number of obs: 4289, groups: HNDid, 1948
##
## Fixed effects:
## Estimate Std. Error df t value Pr(>|t|)
## (Intercept) 9.479e+01 6.474e-01 1.502e+03 146.417 < 2e-16 ***
## CES 5.481e-02 3.562e-02 1.513e+03 1.539 0.124132
## cenAge 2.126e-01 5.570e-02 3.040e+03 3.817 0.000138 ***
## cenSex 3.469e+00 7.135e-01 1.999e+03 4.862 1.25e-06 ***
## cenRace -7.875e-01 7.253e-01 2.025e+03 -1.086 0.277763
## cenPovStat -9.927e-01 7.617e-01 2.007e+03 -1.303 0.192633
## cenEducnum -3.327e-01 1.454e-01 2.019e+03 -2.289 0.022194 *
## CES:cenAge 8.197e-03 3.158e-03 3.223e+03 2.596 0.009476 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Correlation of Fixed Effects:
## (Intr) CES cenAge cenSex cenRac cnPvSt cnEdcn
## CES -0.794
## cenAge 0.391 -0.323
## cenSex -0.041 0.089 -0.011
## cenRace -0.039 0.043 -0.004 -0.015
## cenPovStat 0.085 -0.107 0.035 0.045 -0.171
## cenEducnum -0.168 0.161 0.002 0.047 0.051 0.231
## CES:cenAge -0.308 0.432 -0.787 0.009 0.006 -0.007 0.009This effect remains significant after adjustment for diabetes-related risk factors, even with a lower n (not shown here).
Plotting the effect shows that greater depression is associated with the steepest increase in glucose over time 