Skip to main content
. 2013 Oct 7;8(10):e78292. doi: 10.1371/journal.pone.0078292

Table 3. Comparison between depressed subjects with MDD only, versus those with primary PTSD and comorbid MDDand healthy controls (HC).

Parameters Estimate SEM t p 95% CI boot
(bias) Lower Limit Upper Limit
Early Discounting (β)
HC 0.664 (0.007) 0.134 4.974 <0.001 0.4101 0.9967
MDD only 1.370 (0.053) 0.297 4.613 <0.001 0.862 2.374
MDD + PTSD 1.800 (0.498) 0.442 4.072 <0.001 1.002 8.652
Group differences in early discounting (β* g)
HC vs. MDD only 0.706 (0.005) 0.326 2.168 0.031 0.1028 1.6968
HC vs. MDD + PTSD 1.136 (0.490) 0.462 2.460 0.014 0.265 8.065
MDD only vs. MDD + PTSD 0.430 (0.445) 0.533 0.807 0.420 -0.7816 7.0652
Late Discounting over gains (δ)
HC -0.012 (0.001) 0.013 1.752 0.081 -0.0336 0.0123
MDD only 0.030 (0.001) 0.022 0.507 0.612 0.0006 0.0632
MDD + PTSD 0.057 (0.002) 0.025 3.332 <0.001 0.0126 0.1034
Late Discounting over losses (δ* l)
HC -0.034 (-0.001) 0.008 -4.431 <0.001 -0.0502 0.004
MDD only 0.019 (-0.000) 0.014 1.369 0.172 -0.0558 0.0352
MDD + PTSD -0.027 (-0.002) 0.011 -2.613 0.009 -0.1326 -0.0353
Group differences in late discounting over gains (δ* g)
HC vs. MDD only 0.042 (-0.0002) 0.020 2.132 0.034 0.0035 0.0817
HC vs. MDD + PTSD 0.069 (0.0005) 0.025 2.726 0.007 0.0194 0.1233
MDD only vs. MDD + PTSD 0.027 (0.0007) 0.029 0.932 0.352 -0.0285 0.0835
Group differences in late discounting over losses (δ* g* l)
HC vs. MDD only 0.011 (0.0005) 0.025 0.441 0.659 -0.0417 0.0652
HC vs. MDD + PTSD -0.062 (-0.001) 0.028 -2.191 0.029 -0.1179 -0.0069
MDD only vs. MDD + PTSD -0.073 (0.001) 0.034 -2.189 0.029 -0.1400 -0.0071

Results from best fitting quasi-hyperbolic model and bootstrap analysis showing parameter estimates for each group and group differences in early and late discounting over gains and losses.

β: early discounting, δ: late discounting, g: effect of group on discounting slope, l: effect of loss on discounting slope. Parameter estimates are shown together with their estimated bias based on bootstrap analyses (in brackets), respective standard errors (SEM), t statistic, p values, and 95% confidence intervals from confirmatory nonparametric bootstrap analyses.

Parameter estimates reflect quasi-hyperbolic intertemporal choice functions for each group (β, δ and, δ * l). Interactions with the group dummy (g) reflect differences in slopes between groups for delays smaller than one year (β* g), as well as delays greater than one year in the domain of gains (δ* g) and losses (δ* g * l).