Table 3 Bayesian central estimates of slope coefficients of single-level and multilevel models.

Model Condit. distribution	(4) Normal	(5) NB	(6) NB	(7) NB
TA	0.71	0.08	0.12	0.19
TA	(0.30)	(0.12)	(0.13)	(0.12)
DL	0.08	0.02	0.01	0.01
DL	(0.02)	(0.01)	(0.02)	(0.02)
PA	\(-0.47\)	\(-0.04\)	\(-0.03\)	\(-0.03\)
PA	(0.31)	(0.16)	(0.17)	(0.16)
\(\overline{TA}_i\)			\(-11.6\)	\(-11.13\)
\(\overline{TA}_i\)			(2.37)	(2.23)
\(\overline{DL}_i\)			0.37	0.36
\(\overline{DL}_i\)			(0.19)	(0.18)
\(\overline{PA}_i\)			1.64	1.85
\(\overline{PA}_i\)			(3.07)	(3.07)
\(\overline{TA}_{my}\)				0.57
\(\overline{TA}_{my}\)				(0.29)
\(\overline{DL}_{my}\)				\(-0.07\)
\(\overline{DL}_{my}\)				(0.04)
\(\overline{C}_{my}\)				0.01
\(\overline{C}_{my}\)				(0.00)
Region	FEs	FEs	Pooled	Pooled
Month	FEs	FEs	FEs	Pooled
N	2808	2808	2808	2808
ELPD	\(-8090.2\)	\(-2861.7\)	\(-2833.8\)	\(-2809.8\)
\(\text {ELPD}_{\text {diff}}\)	\(-5280.3\)	\(-51.8\)	\(-24\)	0
SE[\(\text {ELPD}_{\text {diff}}\)]	186.4	13.8	9.5	0

For each regressor, the table display summaries of its marginal posterior distribution: the distribution’s median (top sub-row) and an estimate of the distribution’s standard deviation (bottom sub-row, in parentheses) based on a scaling of the median absolute deviation around that median. \(\text {ELPD}_{\text {diff}}\) corresponds to the difference in expected log predictive density (ELPD) between models, and SE [\(\text {ELPD}_{\text {diff}}\)] to the standard error of that difference, where the reference is the model with the largest ELPD (model (7)).

Quick links

Search