1. Load the Libraries and The Dataset

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

## Loading required package: grid

## Loading required package: Matrix

## Loading required package: survival

## 
## Attaching package: 'survey'

## The following object is masked from 'package:graphics':
## 
##     dotchart

## #StandWithUkraine

## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.2 ──
## ✔ ggplot2 3.4.0     ✔ purrr   0.3.5
## ✔ tibble  3.1.8     ✔ stringr 1.5.0
## ✔ tidyr   1.2.1     ✔ forcats 0.5.2
## ✔ readr   2.1.3     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ tidyr::expand() masks Matrix::expand()
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ✖ tidyr::pack()   masks Matrix::pack()
## ✖ tidyr::unpack() masks Matrix::unpack()

Load the Dataset.
This is the data that I constructed from assignment 3.
I’ll be Working with my data, not what was downloaded from Laulima.
So, I hope I did assignment 3 correctly!!!

mydatasf1<- readRDS("C:/Users/Sferg/Desktop/JABSOM Grad School/Secondary Data Analysis/Week 4/Assignment 4 Fresh start/updated_copy of nhanes_assignment3_finished_v2.rds")

2. Recode the Variables as Factors

Must re-label our codes into readable “factors” so that we can eventually pull it all together to make a single subject characteristic table.

Don’t want our “Sex” variable to read “1”. We want it to say “Male” when we compile it into a table.
This applies to every categorical variable.

Using the mutate function, we don’t overwrite old columns
Instead, we create knew ones. For example: the column riagendr is not overwritten and relabeled by Sex. We create a new column Sex that is derived from the riagendr column

Using the mutate() function with a pipeline is a more recommended approach when working with the tidyverse, as it promotes non-destructive data manipulation and makes the code more readable and maintainable.

mydatasf1 <- mydatasf1 %>%
  mutate(Sex = factor(riagendr, levels = c(1, 2), labels = c("Male", "Female")),
         Age_Group = factor(Age, levels = c(1,2), labels= c("20-64 Years", "≥65 Years")),
         Race = factor(ridreth3, levels = c(1:5, 7), labels = c("Mexican American", "Other Hispanic", 
                                                                "White, Non-Hispanic", "Black,
                                                                Non-Hispanic", "Asian, Non-Hispanic",
                                                                "Other Race")),
         Country_of_Birth = factor(dmdborn4, levels = c(1, 2), labels = c("Born in USA", "Born in Other Country")),
         Marital_Status = factor(dmdmartz, levels = c(1:3), labels = c("Married or Cohabitating",
                                                                           "Divorced/separated/widowed", "Single")),
         Smoking_Status_Group = factor(Smoking_Status, levels = c(1:3), labels = c("Current Smoker", 
                                                                                   "Former Smoker", 
                                                                                   "Never Smoked")),
         Weight = wtmecprp,
         Cluster = sdmvpsu,
         Stratum = sdmvstra)

After this chunk of code is run, notice how mydatasf changes from 15 to 24 variables.

table(mydatasf1$Sex, useNA="ifany")

## 
##   Male Female 
##   7721   7839

table(mydatasf1$Age_Group, useNA="ifany")

## 
## 20-64 Years   ≥65 Years        <NA> 
##        6821        2411        6328

table(mydatasf1$Race, useNA="ifany")

## 
##                                                                     Mexican American 
##                                                                                 1990 
##                                                                       Other Hispanic 
##                                                                                 1544 
##                                                                  White, Non-Hispanic 
##                                                                                 5271 
## Black,\n                                                                Non-Hispanic 
##                                                                                 4098 
##                                                                  Asian, Non-Hispanic 
##                                                                                    0 
##                                                                           Other Race 
##                                                                                 1019 
##                                                                                 <NA> 
##                                                                                 1638

table(mydatasf1$Country_of_Birth, useNA="ifany")

## 
##           Born in USA Born in Other Country                  <NA> 
##                 12525                  3028                     7

table(mydatasf1$Marital_Status, useNA="ifany")

## 
##    Married or Cohabitating Divorced/separated/widowed 
##                       5279                       2148 
##                     Single                       <NA> 
##                       1795                       6338

table(mydatasf1$Smoking_Status_Group, useNA="ifany")

## 
## Current Smoker  Former Smoker   Never Smoked           <NA> 
##           1684           2205           5799           5872

mydatasf1 <- mydatasf1 %>%
  mutate(Race = factor(ridreth3, levels = c(1:4,6,7), labels = c("Mexican American", "Other Hispanic", 
                                                                "White, Non-Hispanic", "Black, Non-Hispanic", 
                                                                "Asian, Non-Hispanic", "Other Race")))
table(mydatasf1$Race, useNA="ifany")

## 
##    Mexican American      Other Hispanic White, Non-Hispanic Black, Non-Hispanic 
##                1990                1544                5271                4098 
## Asian, Non-Hispanic          Other Race 
##                1638                1019

# Reorder the factor levels of the Race variable according to the desired order
mydatasf1$Race <- factor(mydatasf1$Race, levels = c("White, Non-Hispanic", "Black, Non-Hispanic", "Mexican American", "Other Hispanic", "Asian, Non-Hispanic", "Other Race"))

# Create the table with the sorted order
table(mydatasf1$Race, useNA = "ifany")

## 
## White, Non-Hispanic Black, Non-Hispanic    Mexican American      Other Hispanic 
##                5271                4098                1990                1544 
## Asian, Non-Hispanic          Other Race 
##                1638                1019

3. Specify the Survey Design

It is important to Specify the Survey Design BEFORE subsetting the data when complex sampling is involved

I almost finished this entire assignment before realizing I had subsetted before defining survey design. Went back and fixed it all. It took awhile to realize I could no longer use filter() and that I had to switch to subset().

Specifying survey design is a crucial step in analyzing data from complex surveys. Complex surveys typically involve elements such as stratification, clustering, and unequal probability sampling to ensure that the survey results are representative of the target population. These elements can have a significant impact on the accuracy of your statistical inferences, and ignoring them can lead to biased results.

When you specify the survey design, you provide the necessary information about these elements to your statistical software so that it can correctly analyze the data and take into account the survey’s structure. Some of the key components you need to specify when defining the survey design include:

Sampling weights: Sampling weights adjust for the unequal probability of selection that can arise in complex surveys. They are used to account for the survey design and ensure that the results are representative of the target population.

Stratification: Stratification is a technique used to divide the population into homogeneous groups or strata before sampling. Specifying the strata in the survey design allows the statistical software to account for the stratification when calculating variance estimates and other summary statistics.

Clustering: Clustering is a common feature in complex surveys, where units are selected in groups (e.g., households within a neighborhood). Clustering can lead to correlated responses within groups and can affect the precision of your estimates. Specifying the clusters in the survey design allows the statistical software to account for this correlation when analyzing the data.

By specifying the survey design, you ensure that your statistical analyses account for the complex survey structure, which is essential for obtaining unbiased estimates and accurate standard errors. Most statistical software, including R and Python, have dedicated packages for analyzing complex survey data that allow you to specify the survey design and conduct the appropriate analyses. [GPT-4]

Including ^^^ for my own reference later

Use svydesign within the {survey} package

Create survey design objects that include information about sampling weights, stratification, and clustering, which are essential for correctly analyzing complex survey data.

my_data2_svy_design <- svydesign(ids = ~Cluster, 
                    weights = ~Weight, 
                    strata = ~Stratum,
                    nest = TRUE, 
                    data = mydatasf1 %>%
                             select(Sex, Age_Group, Race, Country_of_Birth, Marital_Status,
                                    Smoking_Status_Group, lbxbse, Cluster, Weight, Stratum))

4. Filter the Dataset based on the required conditions (MEC weight, blood selenium, age restrictions)

mydata3_filtered_svy_design <- subset(my_data2_svy_design, Weight > 0 & lbxbse > 0 & !is.na(Age_Group))

5. Create the Subject Characteristic Table

library(gt)
weighted_sub_char_tbl <-tbl_svysummary(mydata3_filtered_svy_design,
                            include=c(Sex, Age_Group, Race, Country_of_Birth, Marital_Status, Smoking_Status_Group),
                            statistic=list(all_categorical()~
                                             "{n_unweighted} ({p}%)"),
                                           digits=list(all_categorical() ~c(0,1)),
                                           missing="no",
                            label = c(Sex = "Sex",
                                      Age_Group = "Age",
                                      Race = "Race/Ethnicity",
                                      Country_of_Birth = "Country of Birth",
                                      Marital_Status = "Marital Status",
                                      Smoking_Status_Group = "Smoking Status")) %>%
  as_gt() %>% 
  tab_header(title = "Table 1. Subject Characteristics")
weighted_sub_char_tbl

1. Create a summary table for weighted data (tbl_svysummary)

## Weighted
tbl_svy2 <- tbl_svysummary(
  mydata3_filtered_svy_design, 
  include=c(Sex, Age_Group, Race, Country_of_Birth, Marital_Status, Smoking_Status_Group), by=lbxbse,
  type = all_continuous() ~ "continuous2",
  statistic=list(all_continuous() ~ c("{mean} ({sd})")),
  missing="no",
  label=list(lbxbse = "Blood selenium (ug/L)"))

tbl_svy2