--- title: "Account for NHANES Sampling Design in Regression" author: "Vy Nguyen" date: "1/9/2023" output: html_document --- ## Goal We will learn how to 1. merge datasets together using tidyverse syntax by using the %>% (AKA pipe) operator 2. run a linear regression model that accounts for NHANES sampling design 3. run a cox proportional hazard model that accounts for NHANES sampling design ```{r setup, include = FALSE} knitr::opts_chunk$set(echo = TRUE) ``` Note that the `echo = FALSE` parameter was added to the code chunk to prevent printing of the R code that generated the plot. ## Install packages to run analysis You only need to run this ONCE! ```{r, echo = FALSE} install.packages("survey") # Needed to run a regression model that accounts for the NHANES sampling design install.packages("tidyverse") # Needed to pipe functions together install.packages("broom") # Needed to easily extract regression statistics ``` ## Load packages Run this to have access to the functions in these packages. ```{r, echo = FALSE} library("survey") library("tidyverse") library("broom") ``` ## Loading NHANES datasets Make sure you set your working directory to be the folder containing the file for the NHANES data. ```{r, echo = FALSE} load("./w - nhanes_1988_2018.RData") ``` ## Merge NHANES datasets Please use any dataset that ends with "_clean". Let's merge the demographics, mortality, and weights datasets together using tidyverse syntax. ```{r, echo = FALSE} nhanes_merged <- full_join(demographics_clean, mortality_clean, by = c("SEQN", "SEQN_new", "SDDSRVYR")) %>% full_join(., # Mean the previous data frame, so the merged dataset of demographic and mortality variables weights_clean, by = c("SEQN", "SEQN_new", "SDDSRVYR")) %>% full_join(., # Mean the previous data frame, so the merged dataset of demographic, mortality, and response variables response_clean, by = c("SEQN", "SEQN_new", "SDDSRVYR")) ``` ## Data dictionary Take a look at the dictionary to know the codename and description of the variables ```{r, echo = FALSE} View(df_dictionary) ``` ## Ensure that sex and race/ethnicity variables are categorical factor() is a base R function that encode a variable as categorical relevel() is a base R function that set the reference group We ensured that sex (RIAGENDR) is a categorical variable. We ensured that race/ethnicity (RIDRETH1) is a categorical variable with the reference group as Non-Hispanic Whites (3). ```{r, echo = FALSE} nhanes_merged <- nhanes_merged %>% mutate(RIAGENDR = factor(RIAGENDR)) %>% mutate(RIDRETH1 = factor(RIDRETH1) %>% relevel(., ref = 3)) ``` ## Choose only variables that you will use Let's subset the nhanes_merged dataset to include only variables that we will use and participants who have info on all selected variables. We will create two subsets: one for the linear model and another for the cox proportional model. Notice the different size of these two subsets. Tip: Make your dataset as small as possible and as big as necessary or else you'll be waiting for results for quite some time! select() is a tidyverse function that choose columns na.omit() is a base R function that exclude rows if it's missing any of the selected columns ```{r, echo = FALSE} # For a generalized linear model, we will have the outcome variable be BMI (BMXBMI) and the predictors as age (RIDAGEYR), sex (RIAGENDR), and race/ethnicity (RIDRETH1). nhanes_merged_for_glm <- nhanes_merged %>% select("SEQN", "SEQN_new", "SDDSRVYR", "SDMVPSU", "SDMVSTRA", "WTMEC2YR", "BMXBMI", "RIDAGEYR", "RIAGENDR", "RIDRETH1") %>% na.omit(.) # For the Cox proportional hazard model, we will have the outcome variable be mortality status (MORTSTAT) and time to death or end of the study period (PERMTH_INT) and the predictors as age (RIDAGEYR), sex (RIAGENDR), and race/ethnicity (RIDRETH1). nhanes_merged_for_cox <- nhanes_merged %>% select("SEQN", "SEQN_new", "SDDSRVYR", "SDMVPSU", "SDMVSTRA", "WTMEC2YR", "MORTSTAT", "PERMTH_INT", "RIDAGEYR", "RIAGENDR", "RIDRETH1") %>% na.omit(.) ``` Check out the dimension of these two different subsets. ```{r, echo = FALSE} dim(nhanes_merged_for_glm) dim(nhanes_merged_for_cox) ``` ## Account for NHANES sampling design svydesign() is a survey function that creates an object to specify the sampling design ```{r, echo = FALSE} dsn_glm <- svydesign(ids = ~SDMVPSU, strata = ~SDMVSTRA, weights = ~WTMEC2YR, nest = TRUE, data = nhanes_merged_for_glm) dsn_cox <- svydesign(ids = ~SDMVPSU, strata = ~SDMVSTRA, weights = ~WTMEC2YR, nest = TRUE, data = nhanes_merged_for_cox) ``` ## Linear regression model Let's run a generalized regression model and another one accounting for NHANES sampling design. BMI is the outcome while the predictors are age, sex, and race/ethnicity. ```{r, echo = FALSE} # Run a generalized linear model model_glm <- glm(BMXBMI ~ RIDAGEYR + RIAGENDR + RIDRETH1, data = nhanes_merged_for_glm) # Run a generalized linear model accounting for the NHANES sampling design model_svyglm <- svyglm(BMXBMI ~ RIDAGEYR + RIAGENDR + RIDRETH1, design = dsn_glm) ``` # Looking at the results of the linear model ```{r, echo = FALSE} # Show the regression statistics on the associations between the predictors and outcome tidy(model_glm) tidy(model_svyglm) # Show the regression statistics on the model glance(model_glm) glance(model_svyglm) ``` ## Cox proportional hazard model Let's run a Cox proportional hazard model accounting for NHANES sampling design. Mortality status and time to death or end of the study period are the outcome and the predictors are age, sex, and race/ethnicity. ```{r, echo = FALSE} # Run a cox proportional hazard model model_coxph <- coxph(Surv(time = PERMTH_INT, event = MORTSTAT) ~ RIDAGEYR + RIAGENDR + RIDRETH1, data = nhanes_merged_for_cox) # Run a cox proportional hazard model accounting for NHANES sampling design model_svycoxph <- svycoxph(Surv(time = PERMTH_INT, event = MORTSTAT) ~ RIDAGEYR + RIAGENDR + RIDRETH1, design = dsn_cox) ``` # Looking at the results of the Cox model ```{r, echo = FALSE} # Show the regression statistics on the associations between the predictors and mortality risk tidy(model_coxph) tidy(model_svycoxph) # Extract the hazard ratios tidy(model_coxph, exponentiate = TRUE) tidy(model_svycoxph, exponentiate = TRUE) # Show the regression statistics on the model glance(model_coxph) # Unfortunately glance() doesn't work on a survey model object glance(model_svycoxph) ``` # Manually extracting or calculating the statistics for a Cox model that accounts for the sampling design ```{r, echo = FALSE} glance_svycoxph <- function(model_object) { summary_svycoxph <- summary(model_svycoxph) data.frame(n = summary_svycoxph$n, nevent = summary_svycoxph$nevent, statistic.log = -2*(model_object$ll[1] - model_object$ll[2]), p.value.log = , statistic.sc = , p.value.sc = , statistic.wald = summary_svycoxph$waldtest["test"], p.value.wald = summary_svycoxph$waldtest["pvalue"], r.squared = 1 - exp((2/n)*(model_object$ll[1] - model_object$ll[2])), r.squared.max = 1 - exp(2*model_object$ll[1]/n), concordance = summary_svycoxph$concordance["C"], std.error.concordance = summary_svycoxph$concordance["se(C)"], logLik = model_object$ll[2], df = length(model_coxph$coefficients), AIC = -2*logLik + 2*df, BIC = -2*logLik + log(n)*df, nobs = summary_svycoxph$n) } glance_svycoxph(model_svycoxph) ```