Table 6.

Multivariate models for predicting log sum of dioxin^a concentrations among adolescent boys in Chapaevsk, Russia

Model #	Predictor	Estimate	p-value	R-square value	Adjusted R-square value
1	Consumption of local non-chicken eat (y/n)	0.59	0.017	0.56	0.52
	log (PCB 118) (ng/g lipid)	0.63	<0.001
2	Age (years)	0.40	<0.001	0.56	0.46
	Weeks of gestation	-0.11	0.005
	Consumption of local dairy (y/n)	-0.76	0.037
	Consumption of local non-chicken meat (y/n)	0.90	<0.001
	Monthly family income level (ordinal, low, med, high)	-0.27	0.020

Abbreviations: Log, logarithm base 10

^aSum of dioxin-like compounds includes PCDDs, PCDFs, and coplanar PCBs

N = 26 for model 1; N = 29 for model 2. Missing data for 3 boys on lipid-adjusted PCB 118, and one boy on questionnaire items (medical history, diet).

Model 1: best fit multivariate model including PCB 118 data.

Model 2: best fit multivariate model (forward and backward selection) not including PCB 118.