Use of data from IPUMS NHIS is subject to conditions including that users
should cite the data appropriately. Use command `ipums_conditions()` for more
details.
data<- haven::zap_labels(data)
In this work, I have studied the effect of education, sex and poverty affect the likelihood of adult mortality. I have used IPUMS data (1997-2018) to perform the analysis. Education is measured as less than high school/ higher than high school. Poverty is measured as ratio of family income to poverty threshold being less or greater than 1. Hence, my event variable is adult mortality and my covariates are education, sex, and poverty .
Filtering data (adults + mortality)
data <- data %>%filter(MORTELIG ==1)data <- data %>%filter(AGE ==18) data <- data%>%filter(complete.cases(.))
data$educcat<-car::Recode(data$EDUC, recodes="100:116 ='No HS' ; 200:504='HS or higher' ; else=NA", as.factor=T)data$educcat<-relevel(data$educcat, ref='No HS') data$povertycat<-car::Recode(data$POVERTY, recodes="10:14='Ratio less than 1'; 20:38='Ratio greater than 1'; else=NA", as.factor=T)data$sex<-as.factor(ifelse(data$SEX==1, "Male", "Female"))data <- data%>%filter(complete.cases(.))
Exponential Model
This model suggests men are 3 times more likely to experience mortality compared to female and adults with higher than HS degree are are lower risk of mortality compared to adults with less than HS degree. Finally, adults with family income: poverty threshold ratio less than 1 are at higher risk of mortality.
Single term deletions
Model:
Surv(death_age, d.event) ~ sex + educcat + povertycat * POVERTY
Df AIC LRT Pr(>Chi)
<none> 4577
sex 1 4647 72.7 <2e-16 ***
educcat 1 4578 3.8 0.052 .
povertycat:POVERTY 1 4575 0.0 0.852
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Covariate Mean Coef Rel.Risk S.E. Wald p
sex
Female 0.481 0 1 (reference)
Male 0.519 1.145 3.144 0.146 0.0000
educcat
No HS 0.455 0 1 (reference)
HS or higher 0.545 -0.239 0.788 0.123 0.0530
povertycat
Ratio greater th 0.757 0 1 (reference)
Ratio less than 0.243 -0.020 0.981 1.501 0.9896
POVERTY 26.319 -0.024 0.976 0.012 0.0420
povertycat:POVERTY
Ratio less than 1: -0.023 0.978 0.122 0.8521
Events 270
Total time at risk 545198
Max. log. likelihood -2282.3
LR test statistic 85.13
Degrees of freedom 5
Overall p-value 1.11022e-16
plot(fit.1)lines(fit.haz.sm, col=2)
Weibull Model
This model also suggests adult men higher risk (3.22) to experience mortality compared to female) and adults with higher than HS degree are are lower risk of mortality (0.85) compared to adults with less than HS degree. Finally, adults with family income: poverty threshold ratio less than 1 are at higher risk of mortality (2.439).
Based on the AIC values obtained below fit4 (log-logistic) model is a best fit.
AIC(fit.1)
[1] 4576.53
AIC(fit.2)
[1] 4077.564
##AIC(fit.3)**
AIC(fit.4)
[1] 4075.564
Piece wise model (exponential)
This model also suggests adult men higher risk (3.6) to experience mortality compared to female and adults with higher than HS degree are are lower risk of mortality (0.785) compared to adults with less than HS degree. Finally, adults with family income: poverty threshold ratio less than 1 are at higher risk of mortality (1.049). Piece Wise model had lowest AIC value so, it is the most suitable model to explain the data.
