Missing.Rmd

---
title: "Missing Variables"
author: "Bryon Langford, Matthew Hoctor"
date: "8/6/2021"
output:
  html_document:
    number_sections: no
    theme: lumen
    toc: yes
    toc_float:
      collapsed: yes
      smooth_scroll: no
  pdf_document:
    toc: yes
  word_document:
    toc: yes
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)

if (Sys.info()["sysname"] == "Windows") 
  knitr::opts_chunk$set(engine.path = list(stata = "C:/Program Files/Stata17/StataBE-64")) 
if ((Sys.info()["sysname"] == "Darwin"))
  knitr::opts_chunk$set(engine.path = list(stata = "/Applications/Stata/StataIC"))
```

```{r libraries, include=FALSE}
library(Statamarkdown)
```

# Examining Missing Alcohol

## Creating the missing alcohol variable

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

ta AL_cat AL_miss,m
```

## Al_miss vs MJ

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta MJ AL_miss,col percent
```

## Al_miss vs BP_cat

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta BP_cat AL_miss,col percent
```

## Al_miss vs gndr

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta gndr AL_miss,col percent
```

## Al_miss vs race_eth

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta race_eth AL_miss,col percent
```

## Al_miss vs SMK_cat

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta SMK_cat AL_miss,col percent
```

## Al_miss vs EDUC_cat

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta EDUC_cat AL_miss,col percent
```

## Al_miss vs sddsrvyr

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): ta sddsrvyr AL_miss,col percent
svy, subpop(if include==1): ta sddsrvyr AL_miss,row percent
```

## Al_miss vs HEI2015

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): mean hei2015, over(AL_miss)
```

## Al_miss vs BMI

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): mean bmxbmi, over(AL_miss)
```

## Al_miss vs Age

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): mean ridageyr, over(AL_miss)
```

## Al_miss vs Income

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

svy, subpop(if include==1): mean indfmpir, over(AL_miss)
```

## Results

Those with missing alcohol use data are:
  * less light MJ use
  * slightly more stage 2 htn
  * Slightly more male
  * Very similar race/eth
  * Slightly more moderate-heavy smokers
  * Slightly less educated
  * Disproportionately in 2017-2018 cycle
  * slightly lower HEI2015
  * Similar BMI
  * Similar Age
  * slightly lower income
  
Most importantly: Very similar results with or without those with missing alcohol data.

# New key results 

## T1 alcohol use re-tabulated

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat .=3, gen(ALC_cat)

label define ALC_cat 0 "None-Light" 1 "Moderate" 2 "Heavy" 3 "Missing"
label values ALC_cat ALC_cat
label variable ALC_cat "Alcohol Use Category"

svy, subpop(if include==1): ta ALC_cat MJ_cat, col percent
ta ALC_cat MJ_cat if include==1,m
```

## T3 Model 2 result

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

xi: svy,subpop(if include==1 & AL_miss==0): logit BP_abn i.MJ2 gndr ridageyr, or
```

## T4 Model 2 result

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat 0=0 1=0 2=0 3=0 4=0 .=1, gen(AL_miss)

label define AL_miss 0 "Not Missing" 1 "Missing"
label values AL_miss AL_miss
label variable AL_miss "Alcohol Use Missing"

xi: svy,subpop(if include==1 & AL_miss==0): mlogit BP_cat i.MJ2 gndr ridageyr, rrr
```

# Examining Excluded in Age 20-59

## Age

### Exclusions vs age criteria

Most observations (44k / 70k) were not in the correct age range.

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

ta AGE_exclude include,m
```

### Mean

mean ages are quite similar

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): mean ridageyr, over(include)
```

## Gender

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): ta gndr include, col percent
ta gndr include if AGE_exclude==0,m
```

## Race

Included participants were a bit whiter than excluded.

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): ta race_eth include, col percent
ta race_eth include if AGE_exclude==0,m
```

## Education

Included were slightly more educated; note difference in less than 9th grade education.

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): ta EDUC_cat include, col percent
ta EDUC_cat include if AGE_exclude==0,m
```

## Tobacco Use

Similar

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): ta SMK_cat include, col percent
ta SMK_cat include if AGE_exclude==0,m
```

## BMI

Similar

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): mean bmxbmi, over(include)
```

## Income

Excluded had slightly lower income.

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): mean indfmpir, over(include)
```

## Alcohol

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

recode AL_cat .=3, gen(ALC_cat)

label define ALC_cat 0 "None-Light" 1 "Moderate" 2 "Heavy" 3 "Missing"
label values ALC_cat ALC_cat
label variable ALC_cat "Alcohol Use Category"

svy, subpop(if AGE_exclude==0): ta AL_cat include, col percent
svy, subpop(if AGE_exclude==0): ta ALC_cat include, col percent
ta ALC_cat include if AGE_exclude==0,m
```

## Diet Quality

Similar

```{stata}
use "data\NHANES0518_new.dta", clear
svyset sdmvpsu [pw=wtmec12yr], strata(sdmvstra)

svy, subpop(if AGE_exclude==0): mean hei2015, over(include)
```