data_cleaning.Rmd

---
title: "Data cleaning file"
author: "Mingzhou_Fu"
date: "9/3/2020"
output: html_document
---

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

```{r message=FALSE, warning=FALSE}
rm(list = ls())
library(tidyverse)
library(writexl)
library(haven)   # read in SAS file
library(compareGroups)

raw_data_path = '/Users/Mingzhou/Desktop/AD_Grant/HRS_raw/'
work_data_path = '/Users/Mingzhou/Desktop/AD_Grant/DiaDem_MR/data/'
output_path = '/Users/Mingzhou/Desktop/AD_Grant/DiaDem_MR/output/'

# Original dataset
raw_dataset = 'hrs_democoggen_wide20190813.Rda'
load(paste0(raw_data_path, raw_dataset)) # -- N = 43,232
# Kunkle PGS
load(paste0(raw_data_path, "european_kunkle_pgss.rda"))
load(paste0(raw_data_path, "african_kunkle_pgss.rda"))
# Newly constructed T2DM PGS
europ_pgst2d = read_sas(paste0(work_data_path, 'ea_ph13_t2dm.sas7bdat'))
afric_pgst2d = read_sas(paste0(work_data_path, 'aa_ph13_t2dm.sas7bdat'))
```

# Dataset prepare
## 1) Merge datasets together
```{r message=FALSE, warning=FALSE, paged.print=FALSE}
kunkle_add = 
  hrs %>% 
  left_join(kunkle_afric, by = c('HHID' = 'HHID', 'PN' = 'PN')) %>% 
  left_join(kunkle_europ, by = c('HHID' = 'HHID', 'PN' = 'PN')) %>% 
  # Add an indicator of race
  mutate(gen_ancestry = case_when(
    !is.na(AAPC1_5A) ~ 'African',
    !is.na(eaPC1_5A) ~ 'European'
  )) %>% 
  mutate(PC1_5A = case_when(
    gen_ancestry == "European" ~ eaPC1_5A,
    gen_ancestry == "African" ~ AAPC1_5A
  )) %>% 
  mutate(PC1_5B = case_when(
    gen_ancestry == "European" ~ eaPC1_5B,
    gen_ancestry == "African" ~ AAPC1_5B
  )) %>% 
  mutate(PC1_5C = case_when(
    gen_ancestry == "European" ~ eaPC1_5C,
    gen_ancestry == "African" ~ AAPC1_5C
  )) %>% 
  mutate(PC1_5D = case_when(
    gen_ancestry == "European" ~ eaPC1_5D,
    gen_ancestry == "African" ~ AAPC1_5D
  )) %>% 
  mutate(PC1_5E = case_when(
    gen_ancestry == "European" ~ eaPC1_5E,
    gen_ancestry == "African" ~ AAPC1_5E
  )) 
# N = 43,232

new_pgs_add = 
  kunkle_add %>% 
  left_join(europ_pgst2d, by = c('HHID' = 'hhid', 'PN' = 'pn')) %>% 
  left_join(afric_pgst2d, by = c('HHID' = 'hhid', 'PN' = 'pn'))
# N = 43,232

# Shortened data
shorten_hrs = 
  new_pgs_add %>% 
  mutate(PGS_AD_Kunkle_1 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_1) ~ AAPH13_KunkleNAPOE_PT_1,
    !is.na(EAPH13_KunkleNAPOE_PT_1) ~ EAPH13_KunkleNAPOE_PT_1
  )) %>% 
  mutate(PGS_AD_Kunkle_0_5 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_5) ~ AAPH13_KunkleNAPOE_PT_0_5,
    !is.na(EAPH13_KunkleNAPOE_PT_0_5) ~ EAPH13_KunkleNAPOE_PT_0_5
  )) %>%
  mutate(PGS_AD_Kunkle_0_3 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_3) ~ AAPH13_KunkleNAPOE_PT_0_3,
    !is.na(EAPH13_KunkleNAPOE_PT_0_3) ~ EAPH13_KunkleNAPOE_PT_0_3
  )) %>% 
  mutate(PGS_AD_Kunkle_0_1 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_1) ~ AAPH13_KunkleNAPOE_PT_0_1,
    !is.na(EAPH13_KunkleNAPOE_PT_0_1) ~ EAPH13_KunkleNAPOE_PT_0_1
  )) %>% 
  mutate(PGS_AD_Kunkle_0_05 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_05) ~ AAPH13_KunkleNAPOE_PT_0_05,
    !is.na(EAPH13_KunkleNAPOE_PT_0_05) ~ EAPH13_KunkleNAPOE_PT_0_05
  )) %>% 
  mutate(PGS_AD_Kunkle_0_01 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_01) ~ AAPH13_KunkleNAPOE_PT_0_01,
    !is.na(EAPH13_KunkleNAPOE_PT_0_01) ~ EAPH13_KunkleNAPOE_PT_0_01
  )) %>% 
  mutate(PGS_AD_Kunkle_0_001 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_0_001) ~ AAPH13_KunkleNAPOE_PT_0_001,
    !is.na(EAPH13_KunkleNAPOE_PT_0_001) ~ EAPH13_KunkleNAPOE_PT_0_001
  )) %>% 
  mutate(PGS_AD_Kunkle_5e_08 = case_when(
    !is.na(AAPH13_KunkleNAPOE_PT_5e_08) ~ AAPH13_KunkleNAPOE_PT_5e_08,
    !is.na(EAPH13_KunkleNAPOE_PT_5e_08) ~ EAPH13_KunkleNAPOE_PT_5e_08
  )) %>% 
  mutate(PGS_T2D_hrs = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2D_DIAGRAM12,
    gen_ancestry == "African" ~ AA_PGS3_T2D_DIAGRAM12
  )) %>% 
  mutate(PGS3_T2DM_0_001 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_0_001,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_0_001
  )) %>% 
  mutate(PGS3_T2DM_0_01 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_0_01,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_0_01
  )) %>% 
  mutate(PGS3_T2DM_0_05 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_0_05,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_0_05
  )) %>% 
  mutate(PGS3_T2DM_0_1 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_0_1,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_0_1
  )) %>% 
  mutate(PGS3_T2DM_0_3 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_0_3,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_0_3
  )) %>% 
  mutate(PGS3_T2DM_1 = case_when(
    gen_ancestry == "European" ~ EA_PGS3_T2DM_1,
    gen_ancestry == "African" ~ AA_PGS3_T2DM_1
  )) %>% 
  mutate(PGS_GENCOG = case_when(
    !is.na(AA_PGS3_GENCOG_CHRG15) ~ AA_PGS3_GENCOG_CHRG15,
    !is.na(EA_PGS3_GENCOG_CHRG15) ~ EA_PGS3_GENCOG_CHRG15
  )) %>% 
  select(HHID, PN, STUDY, AD08, AD10, AD12, R9DIABE, R10DIABE, 
         BIRTHYR, GENDER, SCHLYRS, gen_ancestry, MA009, MIWWAVE,
         APOE2010_bin, PGS_AD_Kunkle_1, PGS_AD_Kunkle_0_5, PGS_AD_Kunkle_0_3, PGS_AD_Kunkle_0_1, PGS_AD_Kunkle_0_05, PGS_AD_Kunkle_0_01, PGS_AD_Kunkle_0_001, PGS_AD_Kunkle_5e_08, PGS_GENCOG,
         PGS_T2D_hrs, PGS3_T2DM_0_001, PGS3_T2DM_0_01, PGS3_T2DM_0_05, PGS3_T2DM_0_1, PGS3_T2DM_0_3, PGS3_T2DM_1,
         PC1_5A, PC1_5B, PC1_5C, PC1_5D, PC1_5E,
         smoke10, R10DRINK, R10BMI, STROKE10, R10HIBPE)
# N = 43,232 | 41 vars
```

## 2) Recode some variables
```{r}
summary(shorten_hrs)
full_data =
  shorten_hrs %>% 
  mutate(AD_cat = case_when(
    AD10 == 3 ~ "Normal",
    AD10 == 2 ~ "CIND",
    AD10 == 1 ~ "Dementia"
  )) %>% 
  mutate(AD10 = case_when(
    AD_cat == "Normal" ~ 0,
    AD_cat == "CIND" ~ 1,
    AD_cat == "Dementia" ~ 2
  )) %>% 
  mutate(AD_former_cat = case_when(
    AD08 == 3 ~ "Normal",
    AD08 == 2 ~ "CIND",
    AD08 == 1 ~ "Dementia"
  )) %>% 
  mutate(AD08 = case_when(
    AD_former_cat == "Normal" ~ 0,
    AD_former_cat == "CIND" ~ 1,
    AD_former_cat == "Dementia" ~ 2
  )) %>% 
  mutate(AD_after_cat = case_when(
    AD12 == 3 ~ "Normal",
    AD12 == 2 ~ "CIND",
    AD12 == 1 ~ "Dementia"
  )) %>% 
  mutate(AD12 = case_when(
    AD_after_cat == "Normal" ~ 0,
    AD_after_cat == "CIND" ~ 1,
    AD_after_cat == "Dementia" ~ 2
  )) %>% 
  mutate(diab_cat = case_when(
    R10DIABE == 1 ~ "Yes",
    R10DIABE == 0 ~ "No"
  )) %>% 
  mutate(diab_former_cat = case_when(
    R9DIABE == 1 ~ "Yes",
    R9DIABE == 0 ~ "No"
  )) %>% 
  mutate(age = 2010 - BIRTHYR) %>% 
  mutate(sex = GENDER - 1) %>% 
  mutate(sex_cat = case_when(
    sex == 0 ~ "Male",
    sex == 1 ~ "Female"
  )) %>% 
  mutate(education = SCHLYRS) %>% 
  mutate(proxy_cat = case_when(
    MA009 == 1 ~ 'Self',
    MA009 == 2 | MA009 == 3 ~ 'Proxy'
  )) %>% 
  mutate(stroke_cat = case_when(
    STROKE10 == 1 ~ "Yes",
    STROKE10 == 0 ~ "No"
  )) %>% 
  mutate(APOE2010_cat = case_when(
    APOE2010_bin == 0 ~ "No",
    APOE2010_bin == 1 ~ "Yes"
  )) %>% 
  mutate(HTN_cat = case_when(
    R10HIBPE == 0 ~ "No",
    R10HIBPE == 1 ~ "Yes"
  )) %>% 
  mutate(smoking_cat = case_when(
    smoke10 == 0 ~ "Never",
    smoke10 == 1 ~ "Former",
    smoke10 == 2 ~ "Current"
  )) %>% 
  mutate(drinking_cat = case_when(
    R10DRINK == 0 ~ "Never",
    R10DRINK == 1 ~ "Ever"
  )) %>% 
  select(HHID, PN, STUDY, MIWWAVE, AD08, AD10, AD12, AD_former_cat, AD_cat, AD_after_cat, 
         R9DIABE, R10DIABE, diab_cat, diab_former_cat,
         age, sex, sex_cat, education, MA009, proxy_cat, gen_ancestry, 
         APOE2010_bin, APOE2010_cat, PGS_AD_Kunkle_1, PGS_AD_Kunkle_0_5, PGS_AD_Kunkle_0_3, PGS_AD_Kunkle_0_1, PGS_AD_Kunkle_0_05, PGS_AD_Kunkle_0_01, PGS_AD_Kunkle_0_001, PGS_AD_Kunkle_5e_08, PGS_GENCOG,
         PGS_T2D_hrs, PGS3_T2DM_0_001, PGS3_T2DM_0_01, PGS3_T2DM_0_05, PGS3_T2DM_0_1, PGS3_T2DM_0_3, PGS3_T2DM_1,
         PC1_5A, PC1_5B, PC1_5C, PC1_5D, PC1_5E,
         smoke10, smoking_cat, R10DRINK, drinking_cat, R10BMI, STROKE10, stroke_cat, R10HIBPE, HTN_cat
         )
# N = 43,232 | 53 vars
```

## 3) Sample selection
```{r}
# Wave 2010 data -- N = 23,614
data_2010 = 
  full_data %>% 
  filter(MIWWAVE == 1)
save(data_2010, file = paste0(work_data_path, "data_2010.rda"))

# Complete cognitive, diabetes and genetic data -- N = 10,876
complete_cogen = 
  data_2010 %>% 
  filter(!is.na(PGS_T2D_hrs) & !is.na(AD10) & !is.na(diab_cat) & !is.na(PGS3_T2DM_0_01))

# No missing in primary model -- age, sex, education  - N = 10,851
complete_covar = 
  complete_cogen %>% 
  filter(!is.na(age) & !is.na(sex) & !is.na(education))

# Age between 50 and 90 -- N = 10,371
age_restrict = 
  complete_covar %>% 
  filter(age <= 90 & age >= 50)

# No reversed cognition -- N = 49
reverse_sample_AD =
  age_restrict %>% 
  filter((AD_former_cat == "Dementia" & AD_cat == "Normal") | (AD_cat == "Dementia" & AD_after_cat == "Normal")) 
# No reversed diabetes -- N = 0
reverse_sample_T2D = 
  age_restrict %>% 
  filter(R9DIABE == 1 & R10DIABE == 0)

final_sample = 
  age_restrict %>% 
  anti_join(reverse_sample_AD, by = c('HHID' = 'HHID', 'PN' = 'PN'))
# N = 10,322

save(final_sample, file = paste0(work_data_path, "eligible_all.rda"))
```

## 4) Separate by ancestry
```{r}
europ =
  final_sample %>% 
  dplyr::filter(gen_ancestry == 'European')
# N = 8,433
afric = 
  final_sample %>% 
  dplyr::filter(gen_ancestry == 'African')
# N = 1,889

save(europ, file = paste0(work_data_path, "europ10.rda"))
save(afric, file = paste0(work_data_path, "afric10.rda"))
```