Table 5.

Linear mixed models for predicting log-transformed plasma HDA (μg)^a in 46 automotive spray painters

Covariatesb	nc	Nd	Parameter estimates	REMLe estimates		ICC	P-valuesf
				${\sigma }_{\text{w}}^2$	${\sigma }_{\text{b}}^2$
Intercept	46	108	−1.676	1.77	0.10	0.05	<0.001
Location	33	78	−0.071	1.44	0.34	0.19	0.806
Coveralls	32	73	−0.423	1.55	0.26	0.14	0.114
Gloves	39	86	−0.096	1.80	0.09	0.05	0.758
Respirator type
Air purifying (half)	35	80	−0.407	1.48	0.12	0.08	0.137
Air purifying (full)	1	3	0.058	2.40	NA	NA	0.942
Supplied air (full or hood)	8	17	−0.034	3.25	0.11	0.03	0.934
PAPR (full or hood)	4	7	0.931	0.66	0.19	0.22	0.068
None	1	1	1.452	NA	NA	NA	0.258
Booth type
Downdraft	31	72	−0.745	1.80	0.02	0.01	0.004
Semi-downdraft	10	21	0.774	1.86	0.03	0.02	0.012
Crossdraft	6	15	0.377	1.04	0.18	0.15	0.310

REML, restricted maximum likelihood; ICC, intraclass correlation coefficient; NA, not applicable.

^aMultiple imputation (n = 10 imputed datasets) was used to impute air-sampling data below MDL.

^bEach fixed-effect dichotomous covariate tested in the linear mixed model for log-transformed plasma HDA level (dependent variable); location (NC = 0, WA = 1), coveralls (no = 0, yes = 1), gloves (no = 0, yes = 1), respirator type (no = 0, yes = 1 for each type) and booth type (no = 0, yes = 1 for each type).

^cNumber of workers.

^dNumber of samples.

^eRestricted maximum likelihood estimates of the log-transformed data for the within-worker (${\sigma }_{\text{w}}^2$) and between-worker variance (${\sigma }_b^2$).

^fP-values are based on F-tests of fixed effects.