Skip to main content

Table 4 Association between alter characteristics and use of contraception using generalized linear model with Poisson regression and robust standard errors (alter-level analysis)

From: The association between men’s family planning networks and contraceptive use among their female partners: an egocentric network study in Madagascar

  Dependent variable: Couple currently using modern contraception
  RR RR RR
Alter characteristics (1) (2) (3)
Female 1.04   
  (0.26)   
Age (years) 0.99   
  (0.00)   
Supportive of contraceptive use   1.12e+ 07***  
   (5.30e+ 06)  
Type of relation
 Provider (nurse, mid-wife, doctor) (Reference group)   
 Community health worker    1.31
    (0.23)
 Health educator    0.83
    (0.10)
 Friend    1.22
    (0.25)
 Partner    1.33**
    (0.14)
 Mother/aunt    1.02
    (0.20)
 Sibling    1.55**
    (0.24)
 Other    0.76
    (0.32)
Age (years) 1.00 1.00 1.00
  (0.01) (0.01) (0.01)
Household size 1.00 0.96 0.98
  (0.03) (0.02) (0.03)
Currently married 1.11 1.21** 1.11
  (0.08) (0.08) (0.15)
Completed primary school education 0.82* 0.85*** 0.91
  (0.08) (0.03) (0.14)
Monthly household earnings (USD) 1.00* 1.00 1.00
  (0.00) (0.00) (0.00)
Sub-district 1 (reference group)
Sub-district 3 0.00*** 0.00*** 0.00***
  (0.00) (0.00) (0.00)
Sub-district 4 0.96 0.94 1.02
  (0.08) (0.04) (0.07)
Sub-district 5 0.94 1.05 0.86
  (0.07) (0.05) (0.09)
Sub-district 6 0.71*** 0.90 0.71***
  (0.07) (0.05) (0.05)
Sub-district 7 0.86* 0.99 0.76**
  (0.05) (0.04) (0.07)
Sub-district 8 0.54*** 1.22*** 0.60***
  (0.06) (0.05) (0.04)
Sub-district 9 0.58*** 0.73** 0.38***
  (0.04) (0.08) (0.04)
Sub-district 10 0.71*** 0.85** 0.65***
  (0.04) (0.04) (0.05)
Observations 143 156 159
Mean of dependent variable 0.69 0.71 0.69
  1. Note: This analysis was restricted to men who had a FP network. Each column presents the findings of a separate regression model. Covariates included in each of these 4 regression models were age, household size, marital status, primary school completion, and household earnings. We also included sub-district fixed effects. Sub-district 2 is not included in this analysis due to missing data from the 5 respondents in that sub-district which had reported an alter but for whom specific alter characteristics are missing likely due to data collection error. We used cluster robust standard errors at the sub-district level. RR Relative risk ratio