Estimates of marginal effects on the probability of tobacco use from the multinomial logit regression of tobacco use status by type and mix of tobacco products in Bangladesh, 2009–2012.

Source: Estimated from ITC Bangladesh Enumeration, 2009 and 2012 (S1 File).Notes1. Pseudo R2 = 0.2671.2. *** and ** indicate significant at 1% and 5% levels respectively.3. The number of observations is 192,449.4. The base categories include the male national sample in low socio-economic status, living in urban areas, in year 2009. The marginal effects on probability represent the change in probability from a discrete change from the base category.5. The base outcome is no use of tobacco products.6. The estimates of the fixed effects for the upazilas/thanas are suppressed.7. The standard errors of estimated coefficients account for the complex sampling design, in particular the household clusters at the village level.Estimates of marginal effects on the probability of tobacco use from the multinomial logit regression of tobacco use status by type and mix of tobacco products in Bangladesh, 2009–2012.