Skip to main content
. 2016 Apr 5;71(5):947–958. doi: 10.1093/geronb/gbw041

Table 4.

Multilevel Logistic Regression Analysis (MCMC) of PSI (Country-Level Coefficients)

Model OR CI (95 %) pMCMC ICC
M0: Empty model 0.041
M1: Individual-level predictors only 0.027
M2a: Intergenerational cohabitation/10 0.69 0.48–0.99 0.054 0.017
M2b: Social trust 1.07 0.75–1.55 0.683 0.030
M2c: Religious population/10 0.84 0.68–1.03 0.092 0.021
M2d: Burdensome eldersa 1.12 0.90–1.40 0.266 0.013
M2e: Health care expenditure 1.14 0.95–1.35 0.141 0.022
M2f: Utilization antidepressants/10 1.01 0.99–1.03 0.376 0.026
M2g: Approval suicide/euthanasia 1.44 1.08–1.91 0.030 0.021

Notes. SHARE, waves 4 (1.1.1) and 5 (1.0), unweighted data, n = 6,791. MCMC: 55,000 iterations, burn-in = 5,000, thinning 20 = eff. sample size ~ 2,500 per chain. OR = adjusted odds ratio of posterior mean, CI = 95 % credible intervals, pMCMC = Bayesian p-value, ICC = intraclass-correlation coefficient. M1 includes all individual-level variables, M2a-M2g include all individual-level variables and one country-level variable each.

aData not available for Italy, model M2g refers to 11 remaining countries. ICC for the reduced reference model M1noIT = 0.013.