preliminary_analysis.Rmd

---
title: "Modeling Food and Housing Insecurity at UTEP - Phase 1"
subtitle: "Data Preprocessing and Exploration"
author: 
 - "George E. Quaye"
 - "John Koomson"
 - "Willliam O. Agyapong"
  
affiliation: "Department of Mathematical Sciences, University of Texas at El Paso"
date: \center University of Texas, El Paso (UTEP)\center
       \center Department of Mathematical Sciences \center
output:
  html_document:
    fig_caption: yes
    keep_tex: no
    number_sections: yes
    toc: yes
    toc_depth: 4
  bookdown::pdf_document2:
    fig_caption: yes
    keep_tex: yes
    latex_engine: pdflatex
    number_sections: yes
    toc: yes
    toc_depth: 4
header-includes:
  - \usepackage{float}
  - \usepackage{setspace}
  - \doublespacing
  - \usepackage{bm}
  - \usepackage{amsmath}
  - \usepackage{amssymb}
  - \usepackage{amsfonts}
  - \usepackage{amsthm}
  - \usepackage{fancyhdr}
  - \pagestyle{fancy}
  - \fancyhf{}
  - \rhead{George Quaye, Johnson Koomson, William Agyapong}
  - \lhead{Analysis Plan - DS 6335}
  - \cfoot{\thepage}
  - \usepackage{algorithm}
  - \usepackage[noend]{algpseudocode}
geometry: margin = 0.8in
fontsize: 10pt
bibliography: references.bib
link-citations: yes
linkcolor: blue
csl: apa-6th-edition-no-ampersand.csl
nocite: |
---
\end{center}
---

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

# Load required packages
library(dplyr)
library(stringr)
library(ggplot2)
library(knitr)
```


```{r data-reading}
file_path <- dirname(rstudioapi::getSourceEditorContext()$path)
FIHI <- read.csv(paste0(file_path, '/datcsv.csv'))


#=========================================================
# Helper functions
#========================================================

to_zero <- function(x) 
{
  return(ifelse(is.na(x), 0, x ))
}

# This function helps restore back any prior missingness. Care must be taken in 
# order not to apply it to a variable where 0 has a different meaning.
to_na <- function(x) 
{
  return(ifelse(x == 0, NA, x ))
}

```



```{r data-preprocessing}

FIHI_new <- FIHI %>% 
  # change column names to lowercase
  rename_with(tolower) %>%
  # Removing unwanted variables
  select(-externalid, -satellite, -email, -datetime, -carddata,
                        -starts_with(c("q5", "q8", "q18", "q21", "q29", "q24", "q36", "q37", "q38", "q39"))
         ) %>%
  # Cleaning column names by removing anything after the first occurrence of a dot in the column name
  rename_with(~gsub('\\..*', '', .x))  

# col_names <- names(FIHI)
# str_view_all(col_names, "\\..*")

## Consolidating common variables split across levels:
## We first create dummy columns for the number of non-na values within each range for the targeted variable.
## Next, we filter out all rows where the count for each of the targeted variables is more than 1.
FIHI_new <- FIHI_new %>% 
  # Creating dummy columns for the number of non-na values within each range for the targeted variable
  mutate(q6_n = rowSums(!is.na(select(., starts_with("q6")))),
         q7_n = rowSums(!is.na(select(., starts_with("q7")))),
         q11_n = rowSums(!is.na(select(., starts_with("q11")))),
         q25_n = rowSums(!is.na(select(., starts_with("q25"))))) 


# Identifyig multiracial instances
FIHI_new_sub <- FIHI_new %>% 
  filter(if_all(q6_n:q25_n, ~ !(.x>1))) %>% # filter out where there are multiple responses for the same question
  select(-(q6_n:q25_n)) %>%
  mutate(across(starts_with(c("q6", "q7", "q11", "q25")), to_zero)) %>%
  mutate(q6 = rowSums(across(starts_with("q6"))),
         q7 = rowSums(across(starts_with("q7"))),
         q11 = rowSums(across(starts_with("q11"))),
         q25 = rowSums(across(starts_with("q25"))),
         .after = starts_with("q4"),
         .keep = "unused") %>%
  relocate(q11, .after = starts_with("q10")) %>%
  relocate(q25, .after = starts_with("q23")) %>%
  mutate(across(starts_with(c("q6", "q7", "q11", "q25")), to_na), 
         across(!respondentid, as.factor) # convert all variables except the respondent id to a factor
         )
  

# Rename the remaining variables
var_names <- c("respondent_id", "enrollment", "employment", "employment_type", "weekly_work_hrs", "ethnicity", "gender", "total_income", "academic_level", "college/school", "mode_transport", "transport_reliability", "living_alone", "have_dependents", "dependents", "household_head", "residence", "permanent_address", "HI", "know_homelessness_studt", "federal_aid", "FI_q26", "FI_q27", "FI_q28", "FI_q30", "FI_q31", "change_expenditures", "change_income", "change_fin_aid", "change_debt")

names(FIHI_new_sub) <- var_names

```

### A glimpse into the issue of multiple responses per individual where a single response was expected.
```{r}
## Issue of multiple race
multi_racial <- FIHI_new %>% 
  filter(q6_n>1) %>%
  select(respondentid, starts_with("q6"))
# names(multi_racial) <- c("respondentid", paste0("q6_", 1:7),"q6_n")
head(multi_racial, 30) %>%
kable(caption = "Individuals identifying with more than one ethnic group")

# Those of more than 1 gender designations.
multi_gender <- FIHI_new %>%
  filter((q7_n>1)) %>%
  select(respondentid, starts_with("q7"))
# names(multi_gender) <- c("RespondentId", paste0("q6_", 1:6),"q7_n")
  kable(multi_gender,caption = "Individuals identifying with more than one gender")
  
# Those who belong to more than 1 college/schools.
multi_colleges <- FIHI_new %>%
  filter((q11_n>1)) %>%
  select(respondentid, starts_with("q11"))
# names(multi_colleges) <- c("RespondentId", paste0("Q11_", 1:9),"q11_n")
 head(multi_colleges, 30) %>%
  kable(caption = "Individuals belonging to more than one college/school")
  
# Those who received more than one source of Federal Aid.
multi_aid <- FIHI_new %>%
  filter((q25_n>1)) %>%
  select(respondentid, starts_with("q25"))
# names(multi_aid) <- c("RespondentId", paste0("q25_", 1:7),"q25_n")
head(multi_aid, 30) %>%
  kable(caption = "Those who received more than one source of Federal Aid")
```


## Missing Data Exploration

### Structure of the preprocessed data
```{r}
glimpse(FIHI_new_sub)
```

### Summary 
```{r}
## Get summary of the data
summary(FIHI_new_sub)
```


### Visualizing missingness 
```{r}
# install.packages("naniar")
# install.packages("UpSetR")
library(naniar)
library(UpSetR)

# explore the patterns
gg_miss_upset(FIHI_new_sub)

# explore missingness in variables with gg_miss_var
miss_var_summary(FIHI_new_sub) %>% kable()

gg_miss_var(FIHI_new_sub) + ylab("Number of missing values")

gg_miss_var(FIHI_new_sub, show_pct = T) 
```


```{r eval=F}

# Permanent Address
plotdf <- FIHI_sub3 %>%
  group_by(permanent_address) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p1 <- ggplot(plotdf, aes(permanent_address,n, fill=as.factor(permanent_address))) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Had permanent address in the past 12 months?", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")


# Spending night elsewhere
plotdf <- FIHI_sub3 %>%
  filter(permanent_address=='No') %>%
  group_by(spent_night_elsewhere) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p2 <- ggplot(plotdf, aes(spent_night_elsewhere,n, fill=as.factor(spent_night_elsewhere))) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="How fequently did you spend the night elsewhere?", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")

# Display plots side by side
p1 + p2


```


## Distribution of Food Insecurity Responses
```{r eval=F}

# Q26: Food bought didn't last
plotdf <- FIHI_sub3 %>%
  group_by(FI_q26) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p1 <- ggplot(plotdf, aes(FI_q26,n, fill=FI_q26)) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Food bought didn't last", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")


# Q27: Balanced diet
plotdf <- FIHI_sub3 %>%
  group_by(FI_q27) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p2 <- ggplot(plotdf, aes(FI_q27,n, fill=FI_q27)) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Couldn't afford balanced meals", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")


# Q27: Balanced diet
plotdf <- FIHI_sub3 %>%
  group_by(FI_q28) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p3 <- ggplot(plotdf, aes(FI_q28,n, fill=FI_q28)) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Ever cut the size of meals?", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")

# Q27: Balanced diet
plotdf <- FIHI_sub3 %>%
  group_by(FI_q30) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p4 <- ggplot(plotdf, aes(FI_q30,n, fill=FI_q30)) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Ever ate less than you should?", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")


# Q27: Balanced diet
plotdf <- FIHI_sub3 %>%
  group_by(FI_q31) %>%
  summarise(n=n()) %>% 
  mutate(pct = n/sum(n),
         lbl = percent(pct))

p5 <- ggplot(plotdf, aes(FI_q31,n, fill=FI_q31)) + 
  geom_bar(stat = "identity",position = "dodge")  +
  geom_text(aes(label = n), size=3, position = position_stack(vjust = 0.5)) +
  scale_fill_brewer(palette="Set2") + 
  labs(x="Ever hungry but didn't eat?", y="Number of Respondents", title="") +
  theme_classic() +
  theme(legend.position = "none")

# Display plots side by side
(p1 + p2)

(p3 + p4)

```
# References{-}

<div id="refs"></div>