H type of relationship with a heavy drinker with responses of `no’ = 0 and `yes’ = 1. We assessed MedChemExpress GW 
501516 sex-specific proportions of reported past-year frequencies (never = 0, occasional = 1?, frequent 5 times) for each type of harm. To calculate unadjusted odds ratios (OR) for reporting greater than the mean number of harm types, we employed binary logistic regression. We conducted separate binary logistic regression analyses by sex, for each of the socio-demographic variables and for respondents’ 
drinking pattern to assess how each variable was independently associated with reporting 5 harm types. To examine associations while holding potential confounders constant, we conducted sex-specific multiple logistic regression (Hosmer and Lemeshow, 2000). We controlled for socio-demographic variables of AMI-1 web theoretical importance (i.e. age group, marital status, rurality and respondents’ drinking) (Benegal et al., 2005; International Institute for Population Sciences and Macro International, 2007). We did not adjust for family income in the main analysis because the real-life value of an Indian rupee varies considerably in urban and rural areas. Missing data that could not be imputed based on responses to other questions were treated as missing in analyses (typically missing was <2 ). Differences were considered statistically significant if P < 0.05.Associations between socio-demographic characteristics, respondents' drinking and reporting harmsIn a multiple logistic regression model that included sex, age group, marital status, rurality and respondents' own drinking, males had approximately 1.5 greater odds of reporting 5 harm types than females (P < 0.001) (Supplementary Table 1). Among all respondents, being age 25?9 or age 40?0, married/cohabitating and living in a rural area also significantly predicted reporting 5 harm types, controlling for other factors. Among females, controlling for other factors, being age 25?9 and being married/cohabitating significantly predicted reporting 5 harm types, but being age 40?0 and living in a rural area was not significantly associated with reporting of harms (Table 4). Additionally, female respondents in the upper family income quartiles had greater odds of reporting 5 harm types compared with the lowest income quartile, controlling for age group, marital status, rurality and respondents' drinking. Those in the second quartile had 1.4 greater odds of reporting 5 harm types (P = 0.034), those in the third income quartile had 1.9 (P < 0.001) and respondents in the highest quartile had 3.4 greater odds, compared with those in the lowest income quartile (P < 0.001). Among males, controlling for other factors, being age 25?9 or age 40?0 and living in a rural area significantly predicted reporting 5 harm types, but being married/cohabitating was not significantly associated with reporting of harms (Table 5). Male respondents in the upper two family income quartiles also had greater odds of reporting 5 harm types compared with the lowest income quartile, and the magnitude was similar to that of females. The odds of reporting greater than the mean number (5) of types of harm from a heavy drinker varied by sex and respondents' drinking patterns (Tables 4 and 5). Among both sexes, differences between abstainers and non-binge drinkers were not significant, controlling forRESULTSTable PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19890549 1 describes the characteristics of those in this study who reported having a heavy drinker in their lives (n = 5375). Nearly 36 were femal.H type of relationship with a heavy drinker with responses of `no’ = 0 and `yes’ = 1. We assessed sex-specific proportions of reported past-year frequencies (never = 0, occasional = 1?, frequent 5 times) for each type of harm. To calculate unadjusted odds ratios (OR) for reporting greater than the mean number of harm types, we employed binary logistic regression. We conducted separate binary logistic regression analyses by sex, for each of the socio-demographic variables and for respondents’ drinking pattern to assess how each variable was independently associated with reporting 5 harm types. To examine associations while holding potential confounders constant, we conducted sex-specific multiple logistic regression (Hosmer and Lemeshow, 2000). We controlled for socio-demographic variables of theoretical importance (i.e. age group, marital status, rurality and respondents’ drinking) (Benegal et al., 2005; International Institute for Population Sciences and Macro International, 2007). We did not adjust for family income in the main analysis because the real-life value of an Indian rupee varies considerably in urban and rural areas. Missing data that could not be imputed based on responses to other questions were treated as missing in analyses (typically missing was <2 ). Differences were considered statistically significant if P < 0.05.Associations between socio-demographic characteristics, respondents' drinking and reporting harmsIn a multiple logistic regression model that included sex, age group, marital status, rurality and respondents' own drinking, males had approximately 1.5 greater odds of reporting 5 harm types than females (P < 0.001) (Supplementary Table 1). Among all respondents, being age 25?9 or age 40?0, married/cohabitating and living in a rural area also significantly predicted reporting 5 harm types, controlling for other factors. Among females, controlling for other factors, being age 25?9 and being married/cohabitating significantly predicted reporting 5 harm types, but being age 40?0 and living in a rural area was not significantly associated with reporting of harms (Table 4). Additionally, female respondents in the upper family income quartiles had greater odds of reporting 5 harm types compared with the lowest income quartile, controlling for age group, marital status, rurality and respondents' drinking. Those in the second quartile had 1.4 greater odds of reporting 5 harm types (P = 0.034), those in the third income quartile had 1.9 (P < 0.001) and respondents in the highest quartile had 3.4 greater odds, compared with those in the lowest income quartile (P < 0.001). Among males, controlling for other factors, being age 25?9 or age 40?0 and living in a rural area significantly predicted reporting 5 harm types, but being married/cohabitating was not significantly associated with reporting of harms (Table 5). Male respondents in the upper two family income quartiles also had greater odds of reporting 5 harm types compared with the lowest income quartile, and the magnitude was similar to that of females. The odds of reporting greater than the mean number (5) of types of harm from a heavy drinker varied by sex and respondents' drinking patterns (Tables 4 and 5). Among both sexes, differences between abstainers and non-binge drinkers were not significant, controlling forRESULTSTable PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19890549 1 describes the characteristics of those in this study who reported having a heavy drinker in their lives (n = 5375). Nearly 36 were femal.