Table 5.
Coefficients of gratitude predicting triglyceride levels in sensitivity analyses.
| Variation in analysis | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| β | p | β | p | β | p | |
| Original results (for comparison) | − .10 | < .001 | − .06 | .007 | − .06 | .028 |
| Alternative data pre-processing choices | ||||||
| Gratitude reflected and log-transformed | .10a | < .001 | .06a | .011 | .05a | .033 |
| Triglyceride level log-transformed (after winsorization at 3 SD) | − .10 | < .001 | − .06 | .013 | − .05 | .058 |
| Winsorization (3 SD) applied to all variables | − .11 | < .001 | − .07 | .004 | − .06 | .017 |
| Winsorization (4 SD) applied to all variables | − .10 | < .001 | − .07 | .004 | − .07 | .013 |
| Sub-sample analyses | ||||||
| Only participants with complete data (i.e., listwise deletion; n = 1704) | − .11 | < .001 | − .07 | .003 | − .07 | .010 |
| Only participants not taking antihyperlipidemic medication (n = 1266) | − .11 | < .001 | − .07 | .022 | − .05 | .089 |
| Only MIDUS 2 data (n = 1054) | − .09 | .002 | − .05 | .124 | − .05 | .175 |
| Only MIDUS Refresher data (n = 746) | − .12 | .002 | − .08 | .022 | − .06 | .130 |
| Adjustment for multiple comparisonsb | ||||||
| Adjustment with Hommel procedure | − .10 | < .001 | − .06 | .050 | − .06 | .168 |
| Adjustment with Bonferroni procedure | − .10 | < .001 | − .06 | .064 | − .06 | .252 |
| Adjustment with Benjamini–Hochberg procedure | − .10 | < .001 | − .06 | .021 | − .06 | .063 |
Bolded p-values indicate statistical significance at the .05 level.
aCoefficients were opposite in sign compared to the rest of the results because gratitude was reflected (i.e., reversed).
bp-value adjustments were conducted taking into account all three dependent variables (triglycerides, LDL, HDL) across the first three models (i.e., a total of nine p-values).