. Author manuscript; available in PMC: 2022 Mar 16.

Published in final edited form as: Environ Sci Technol. 2021 Feb 24;55(6):3786–3795. doi: 10.1021/acs.est.0c02256

Table 2.

Tests of group differences for resampled case study data for censoring at the CCLV

Approach	Distribution	Stove Type 1 (resampled from Eco Chula XXL)		Stove Type 2 (resampled from Butterfly Model)		p-value
Approach	Distribution	n	Mean ± SD (ng/μL)	n	Mean ± SD (ng/μL)	p-value

Uncensored
Full sample	Normal	45	0.139 ± 0.156	40	0.079 ± 0.056	0.023^a
Full sample Maximum Likelihood Estimation	Lognormal	45	0.145 ± 0.225	40	0.089 ± 0.115	0.103^b
After censoring 53 observations^c
Substitute CCLV/2	Normal	45	0.143 ± 0.153	40	0.081 ± 0.049	0.017^a
Maximum Likelihood Estimation	Lognormal	45	0.146 ± 0.200	40	0.092 ± 0.045	0.050^b

Test for difference of means (i.e., that difference of means is zero). Repeating the t-tests on log-transformed DBA data yielded p-value=0.111 for the full sample and p-value=0.051 after censoring.

Test for the difference between the two groups expressed as the ratio of their geometric means

Stove 1: 25 observations were censored; Stove 2: 28 observations were censored