Table 2.
Multilevel models examining associations between perceived friends’ approval and indices of alcohol use.
| Weekly Drinks (DDQ)1 | AUDIT1,2 | Peak-eBAC | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
|
| |||||||||||
| Rate Ratio | CI95% LL | CI95% UL | p-value | Rate Ratio | CI95% LL | CI95% UL | p-value | β | SE | p-value | |
|
|
|
|
|||||||||
| Step 1: Control Variables | |||||||||||
| Cohort (RCT-1=0, RCT-2=1) γ01 | 1.115 | 0.955 | 1.300 | .168 | 1.007 | 0.928 | 1.093 | .866 | .00 | .04 | .999 |
| Campus (Private=0, Public=1) γ02 | 0.804 | 0.687 | 0.940 | .007 | 0.926 | 0.852 | 1.007 | .072 | −.10 | .04 | .014 |
| Sex (M=0, F=1) γ03 | 0.598 | 0.513 | 0.697 | <.001 | 0.784 | 0.722 | 0.851 | <.001 | .01 | .04 | .823 |
| Age γ10 | 0.990 | 0.941 | 1.042 | .704 | 0.986 | 0.958 | 1.015 | .341 | −.11 | .04 | .005 |
| Time (months since baseline) γ20 | 0.992 | 0.985 | 0.999 | .033 | 0.998 | 0.993 | 1.003 | .396 | −.06 | .02 | .009 |
| R2 = .087 | R2 = .096 | R2 = .033 | |||||||||
| Step 2: Main Effects | |||||||||||
| Friends Approval (BW) γ04 | 1.547 | 1.423 | 1.666 | <.001 | 1.235 | 1.180 | 1.292 | <.001 | .26 | 0.04 | <.001 |
| Friends Approval (WI) γ30 | 1.088 | 1.034 | 1.144 | .001 | 1.089 | 1.049 | 1.130 | <.001 | .05 | 0.02 | .015 |
| R2 = .236; ΔR2 = .149 | R2 = .275; ΔR2 = .179 | R2 = .096; ΔR2 = .063 | |||||||||
Note: BW = Between-person level (i.e., aggregated person means). WI = Within-person level (i.e., person-mean centered).
1.
Because weekly drinks and AUDIT scores are skewed count variables, Poisson count regression was used to estimate rate ratios (i.e., RR) representing proportional change for each unit increase in the predictor (e.g., RR of 1.10 = 10% increase for each unit change in the independent variable).
2.
Because the distribution of AUDIT scores did not include any zeros, this model was fit using a truncated-Poisson approach.