dataPreparation.Rmd

---
title: "Datenaufbereitung"
author: "Stefan P. Thoma"
date: "3/2/2020"
output: html_document
---

# Setup

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
if (!require("pacman")) install.packages("pacman")
p_load(tidyverse)
```

# Load Data

```{r}
master <- read_csv("data/Master/masterData.csv")
exp_uebung <- read_csv("data/ExperimentelleUebung/Exp_Data.csv")
```
```{r}
names(master)

names(exp_uebung)
```

```{r}

master <- master %>% 
  mutate(Leiter = "Stefan",
        sex = ifelse(sex == 1, "W", "M"))
exp_uebung <- exp_uebung %>% 
  rename(sex = Geschlecht)

namesInExp <- names(exp_uebung)
namesInMaster <- names(master)
namesInMasterNotInExp <- namesInMaster[!namesInMaster %in% namesInExp]
namesInExpNotInMaster <- namesInExp[!namesInExp %in% namesInMaster]

master[namesInExpNotInMaster] <- NA
exp_uebung[namesInMasterNotInExp] <- NA

namesInMaster <- names(master)
names(exp_uebung)
# make order the same for exp_uebung
exp_uebung <- exp_uebung %>%
  dplyr::select(all_of(namesInMaster))

```

```{r}
data <- bind_rows(exp_uebung, master)
data <- data %>% dplyr::select(-`sex_3_TEXT - Topics`)
```

## Recode Items
### Define Items to recode

We can actually look at the data to see which items are reversely coded.
This should be done only as a confirmation. 
The decision should be made based on the papers introducing the respective scale.
Also, it should be fairly obvious which are reversely coded when reading the questions.

Look what data looks like regarding coding.
Checked with a principal component analysis.
```{r, eval = FALSE}
psych::pca(r = data %>% dplyr::select(starts_with("nep")))
psych::pca(r = data %>% dplyr::select(starts_with("nr")))
psych::pca(r = data %>% dplyr::select(starts_with("ccs")))
psych::pca(r = data %>% dplyr::select(starts_with("ipq")))
psych::pca(r = data %>% dplyr::select(starts_with("sod"))) # looks like sod_1 is kinda different! 
# But according to theory, this is correctly coded. 
```
Define Items to be recoded.
```{r}
# define items to be recoded (as defined in the papers)
nep_reorder <- paste("nep", c(2, 4, 6, 8, 10, 12, 14), sep = "")
nr_reorder <- paste("nr", c(2, 3, 10, 11, 13, 14, 15, 18), sep = "")
# in the ccs, they are all coded the same way. 
ipq_reorder <- paste("ipq", c(3, 9, 11), sep = "") 
sod_reorder <- paste("sod", c( 2, 4, 5, 6, 7, 8), sep = "_")
```

Check on what scale the items are coded:
```{r, EVAL = FALSE}
range(data %>%   dplyr::select(starts_with("nep"))) # 1 to 5 so 1 becomes 5, 2 becomes 4, etc...
range(data %>%   dplyr::select(starts_with("nr"))) # 1 to 5 so 1 becomes 5, 2 becomes 4, etc...
range(data %>%   dplyr::select(starts_with("ccs"))) # 1 to 5 so 1 becomes 5, 2 becomes 4, etc...
summary(data %>% dplyr::select(starts_with("ipq"))) # 1 to 7 so 1 becomes 7, 2 becomes 6, etc...
summary(data %>% dplyr::select(starts_with("sod"))) # 1 to 7  so 1 becomes 7, 2 becomes 6, etc...
```

### helping function
```{r recode}
umpolen7 <- function(items, dat){
  for(i in items){
    
    dat[[i]] <- ifelse(dat[[i]] == 1, 7, 
                       ifelse(dat[[i]] == 2, 6, 
                              ifelse(dat[[i]] == 3, 5, 
                                     ifelse(dat[[i]] == 4, 4, 
                                            ifelse(dat[[i]] == 5, 3, 
                                                   ifelse(dat[[i]] == 6, 2, 
                                                          ifelse(dat[[i]] == 7, 1, NA)))))))
    
  }
  return(dat)
}

umpolen5 <- function(items, dat){
  for(i in items){
    
    dat[[i]] <- ifelse(dat[[i]] == 1, 5, 
                       ifelse(dat[[i]] == 2, 4, 
                              ifelse(dat[[i]] == 3, 3, 
                                     ifelse(dat[[i]] == 4, 2, 
                                            ifelse(dat[[i]] == 5, 1, NA)))))
  }
  return(dat)
}


```
### finally recode
```{r}
data_clean <- umpolen7(c(sod_reorder, ipq_reorder), dat = data)

data_clean <- umpolen5(c(nep_reorder, nr_reorder), dat = data_clean)
```
### did it work?
```{r}
data <- data %>% filter(Leiter != "Stefan") 
psych::pca(r = data_clean %>% dplyr::select(starts_with("nep")))
psych::pca(r = data_clean %>% dplyr::select(starts_with("nr")))
psych::pca(r = data_clean %>% dplyr::select(starts_with("ccs")))
psych::pca(r = data_clean %>% dplyr::select(starts_with("ipq")))
psych::pca(r = data_clean %>% dplyr::select(starts_with("sod")))
```
Again,  it becomes apparent that sod_1 is weird. Go and check the question and think why this could be. 
Should we still keep it in the analysis?

# Calculate score means
I just calculate it for one of the scores. The rest you have to do yourself. 

```{r  score means}
data_clean$nep <- rowMeans(data_clean %>% dplyr::select(starts_with("nep")))
data_clean$nr <-  rowMeans(data_clean %>% dplyr::select(starts_with("nr")))
data_clean$ccs <- rowMeans(data_clean %>% dplyr::select(starts_with("ccs")))
data_clean$ipq <- rowMeans(data_clean %>% dplyr::select(starts_with("ipq")))
data_clean$sod <- rowMeans(data_clean %>% dplyr::select(starts_with("sod")))
data_clean$ses <- rowMeans(data_clean %>% dplyr::select(starts_with("SES")))
```

```{r}
write_csv(data_clean, "data/cleanData.csv")
```