02_decriptive-statistics_2024.Rmd

---
title: "02_descriptive_statistics"
author: "Caroline & Patrycja"
date: "2023-09-05"
output: html_document
---


```{r}
#Clear the environment
rm(list = ls()) 

#Set working directory

```


```{r}
#Loading relevant packages

library(tidyverse)
library(dplyr)
library(psych) #To compute summary of descriptive statistics
library(effectsize) #For computing the effect sizes of the t-tests
library(ggplot2) #For the last plot
library(tidyr) #For the last plot


```


```{r}

#Load the dataset into an object
#This line of code works to load the port data
port_preprocessed_data <- readRDS("flare-kurf-port-data-2023.rds")

#Checking that it works, and loaded the data correctly
view(port_preprocessed_data)

```


```{r}

#Computing the descriptive statistics for the linear regressions.

#Creating a new dataframe with only the relevant variables for the descriptive statistics for the linear regresssion. 
# For reference, this include the means (standard deviations) of the PHQ-9, GAD-7, and WSAS, for the first and last treatment sessions, as well as the change in scores, for each category. 

linear_regression_variables <- port_preprocessed_data[c("reliable_recovery", "reliable_improvement", "recovered", 
                               "delta_wsas", "delta_phq9", "delta_gad7", 
                               "last_treatment_wsas", "first_wsas", 
                               "last_treatment_phq9", "first_phq9", 
                               "last_treatment_gad7", "first_gad7")]

# View the new dataset
View(linear_regression_variables)
```


```{r}

### For the recovered category

# Subset the data to include only those participants who answered 'yes'
recovered_yes <- linear_regression_variables[linear_regression_variables$recovered == "yes", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(recovered_yes)


# Subset the data to include only those participants who answered 'no'
recovered_no <- linear_regression_variables[linear_regression_variables$recovered == "no", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(recovered_no)

```

```{r}

### For the reliable improvement category

# Subset the data to include only those participants who answered 'yes'
reliable_improvement_yes <- linear_regression_variables[linear_regression_variables$reliable_improvement == "yes", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(reliable_improvement_yes)


# Subset the data to include only those participants who answered 'no'
reliable_improvement_no <- linear_regression_variables[linear_regression_variables$reliable_improvement == "no", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(reliable_improvement_no)


```

```{r}

### For the reliable recovery category

# Subset the data to include only those participants who answered 'yes'
reliable_recovery_yes <- linear_regression_variables[linear_regression_variables$reliable_recovery == "yes", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(reliable_recovery_yes)


# Subset the data to include only those participants who answered 'no'
reliable_recovery_no <- linear_regression_variables[linear_regression_variables$reliable_recovery == "no", ]

# Use the descriptives function from the psych package on the subsetted data to obtain table with descriptives
describe(reliable_recovery_no)


```


```{r}

#Computing the descriptive statistics for the participants, including their age, gender, etc. distributions. 
#Viewing the variables to see which ones are necessary for the descriptives.

colnames(port_preprocessed_data)

#The ones necessary are "demographics_age_at_screening_years", "no_treatments_completed", "ieso_diagnosis_name", "demographics_gender", "demographics_employment"

#Creating a new dataframe with those variables.
descriptives_variables <- port_preprocessed_data[c("demographics_age_at_screening_years", "no_treatments_completed", 
                                                   "ieso_diagnosis_name", 
                                                   "demographics_gender", 
                                                   "demographics_employment")]

#Checking it was implemented as intended
View(descriptives_variables)

#Descriptives age and n treatments

describe(descriptives_variables)

# Computing a frequency distribution for the different diagnoses
table(descriptives_variables$ieso_diagnosis_name)

# The majority of the participants were diagnosed with either generalised anxiety disorder (44.4%) or a depressive episode (28.8%). 

# Computing a frequency distribution for the different gender categories
table(descriptives_variables$demographics_gender)


# Computing a frequency distribution for the different types of employment status. 
table(descriptives_variables$demographics_employment)

```

```{r}

linear_regression_variables <- port_preprocessed_data[c("reliable_recovery", "reliable_improvement", "recovered", 
                               "delta_wsas", "delta_phq9", "delta_gad7", 
                               "last_treatment_wsas", "first_wsas", 
                               "last_treatment_phq9", "first_phq9", 
                               "last_treatment_gad7", "first_gad7")]
```


```{r}

#Using indexing to select the correct participants from the broader datasample, such that I can double check the ~1 week estimate between the day where they completed the baseline assessment and the day they started treatment. 

#Loading the initial / archived dataset
archived_dataset <- readRDS("~/Library/CloudStorage/OneDrive-King'sCollegeLondon/KURF/Data Analysis/Data/flare-undergrad-port-data-2023.rds")

#Loading final PORT KURF dataset
final_dataset <- readRDS("~/Library/CloudStorage/OneDrive-King'sCollegeLondon/KURF/Data Analysis/Data/02_preprocessed_port_data.rds")

final_participants_archived_dataset <- archived_dataset[archived_dataset$participant_id %in% final_dataset$participant_id, ]


# Checking if all participant_id values in final_participants_archived_dataset are present in final_dataset
all_ids_present <- all(final_participants_archived_dataset$participant_id %in% final_dataset$participant_id)

# Checking if all participant_id values in final_dataset are present in final_participants_archived_dataset
all_ids_present_in_archived <- all(final_dataset$participant_id %in% final_participants_archived_dataset$participant_id)

# Computing final check using an ifelse argument 
if (all_ids_present && all_ids_present_in_archived) {
  print("YES.")
} else {
  print("NO.")
}

#Response is YES, indicating that the selection of participants was implemented as intended. 


```


```{r}
# Computing the average time between the ieso_assessment_date, and the ieso_first_treatment_date

#Checking the type of variable that the relevant columns are in 
class(final_participants_archived_dataset$ieso_assessment_date)
class(final_participants_archived_dataset$ieso_first_treatment_date)

#They are already in a date format, however, I need them to be in a different type of date format for the function that I am about to use. Therefore, I am converting them to Date objects. 
final_participants_archived_dataset$ieso_assessment_date <- as.Date(final_participants_archived_dataset$ieso_assessment_date)

final_participants_archived_dataset$ieso_first_treatment_date <- as.Date(final_participants_archived_dataset$ieso_first_treatment_date)

#Checking this was implemented as intended
class(final_participants_archived_dataset$ieso_assessment_date)
class(final_participants_archived_dataset$ieso_first_treatment_date)

# Calculate the time difference
time_difference <- difftime(final_participants_archived_dataset$ieso_first_treatment_date, final_participants_archived_dataset$ieso_assessment_date, units = "days")


# Calculate the average time difference
mean_time_difference <- mean(time_difference, na.rm = TRUE)

sd_time_difference <- sd(time_difference, na.rm = TRUE)

# Print the average time difference
print(mean_time_difference)
print(sd_time_difference)


#The mean (sd) number of days between the two assessments is 12.2 (9.2) days. 


```

```{r}
#Computing independent samples t-tests to check whether the differences in symptom severity are significantly different for the yes and no groups in each 

#GAD-7

t.test(port_preprocessed_data$delta_gad7 ~ port_preprocessed_data$recovered, var.equal = TRUE)

cohens_d(delta_gad7 ~ recovered, data = port_preprocessed_data)

t.test(port_preprocessed_data$delta_gad7 ~ port_preprocessed_data$reliable_improvement, var.equal = TRUE)

cohens_d(delta_gad7 ~ reliable_improvement, data = port_preprocessed_data)

t.test(port_preprocessed_data$delta_gad7 ~ port_preprocessed_data$reliable_recovery, var.equal = TRUE)

cohens_d(delta_gad7 ~ reliable_recovery, data = port_preprocessed_data)


```


```{r}
#Computing independent samples t-tests to check whether the differences in symptom severity are significantly different for the yes and no groups in each 

#PHQ-9

t.test(port_preprocessed_data$delta_phq9 ~ port_preprocessed_data$recovered, var.equal = TRUE)

cohens_d(delta_phq9 ~ recovered, data = port_preprocessed_data)


t.test(port_preprocessed_data$delta_phq9 ~ port_preprocessed_data$reliable_improvement, var.equal = TRUE)

cohens_d(delta_phq9 ~ reliable_improvement, data = port_preprocessed_data)


t.test(port_preprocessed_data$delta_phq9 ~ port_preprocessed_data$reliable_recovery, var.equal = TRUE)

cohens_d(delta_phq9 ~ reliable_recovery, data = port_preprocessed_data)


```

```{r}
#Computing independent samples t-tests to check whether the differences in symptom severity are significantly different for the yes and no groups in each 

#WSAS
t.test(port_preprocessed_data$delta_wsas ~ port_preprocessed_data$recovered, var.equal = TRUE)

cohens_d(delta_wsas ~ recovered, data = port_preprocessed_data)


t.test(port_preprocessed_data$delta_wsas ~ port_preprocessed_data$reliable_improvement, var.equal = TRUE)

cohens_d(delta_wsas ~ reliable_improvement, data = port_preprocessed_data)


t.test(port_preprocessed_data$delta_wsas ~ port_preprocessed_data$reliable_recovery, var.equal = TRUE)

cohens_d(delta_wsas ~ reliable_recovery, data = port_preprocessed_data)


```


```{r}

####This section was added after recieving additional variables to add further detail to the paper's manuscript

# Set working directory & Import data 

#Checking data were loaded as intended 
View(additional_kurf_variables)

```

```{r}

# Calculating the mean(SD) number of days betweeen assessment and the first treatment session
mean_difference_assessment_first_treatment <- mean(as.numeric(additional_kurf_variables$ieso_assessment_date - additional_kurf_variables$ieso_first_treatment_date), na.rm = TRUE)

sd_difference_assessment_first_treatment <- sd(as.numeric(additional_kurf_variables$ieso_assessment_date - additional_kurf_variables$ieso_first_treatment_date), na.rm = TRUE)

# Printing the mean difference
print(mean_difference_assessment_first_treatment)
print(sd_difference_assessment_first_treatment)

##The difference is 12.19 (9.20) days 

```


```{r}

# Calculating the mean(SD) number of days betweeen the follow-up assessment and the last treatment session
mean_difference_followup_last_treatment <- mean(as.numeric(additional_kurf_variables$ieso_last_treatment_date - additional_kurf_variables$followup_date), na.rm = TRUE)

sd_difference_followup_last_treatment <- sd(as.numeric(additional_kurf_variables$ieso_last_treatment_date - additional_kurf_variables$followup_date), na.rm = TRUE)

# Printing the mean difference
print(mean_difference_followup_last_treatment)
print(sd_difference_followup_last_treatment)

##The difference is 44.07 (21.23) days 

```

```{r}

# Computing additional descriptive statistics on participant demographics

## Biological sex
frequency_participant_biological_sex <- table(additional_kurf_variables$demographics_biological_sex)

print(frequency_participant_biological_sex)



## Gender
frequency_participant_gender <- table(additional_kurf_variables$demographics_gender)

print(frequency_participant_gender)



##Ethnic origin 
frequency_participant_ethnic_origin <- table(additional_kurf_variables$demographics_ethnic_origin)

print(frequency_participant_ethnic_origin)



## Employment 
frequency_participant_employment <- table(additional_kurf_variables$demographics_employment)

print(frequency_participant_employment)


```

```{r}
#Creating a bar plot to visualise the association between the change scores and the mean levels of improvement 

#Starting by creating a dataframe with the information of the factors to be visualised

improvement_0 <- filter(port_preprocessed_data, retrospectively_reported_improvement == 0) #total of 2 patients
improvement_1 <- filter(port_preprocessed_data, retrospectively_reported_improvement == 1) #total of 3 patients
improvement_2 <- filter(port_preprocessed_data, retrospectively_reported_improvement == 2) #total of 34 patients
improvement_3 <- filter(port_preprocessed_data, retrospectively_reported_improvement == 3) #total of 62 patients
improvement_4 <- filter(port_preprocessed_data, retrospectively_reported_improvement == 4) #total of 34 patients
#Overall 135 patients in the different categories, so everyone is included
```



```{r}
bar_data <- port_preprocessed_data %>%
  group_by(retrospectively_reported_improvement) %>%
  summarise(mean_gad7 = mean(delta_gad7), 
            mean_phq9 = mean(delta_phq9),
            mean_wsas = mean(delta_wsas),
            sd_gad7 = sd(delta_gad7),
            sd_phq9 = sd(delta_phq9),
            sd_wsas = sd(delta_wsas))

# Reshape data to long format for means
long_means <- bar_data %>%
  pivot_longer(cols = starts_with("mean_"), 
               names_to = "change_score_type_mean", 
               names_prefix = "mean_", 
               values_to = "mean")

# Reshape data to long format for standard deviations
long_sds <- bar_data %>%
  pivot_longer(cols = starts_with("sd_"), 
               names_to = "change_score_type_sd", 
               names_prefix = "sd_", 
               values_to = "sd")

# Combine means and standard deviations into a single dataset
long_bar_data_step2 <- long_means %>%
  left_join(long_sds, by = c("retrospectively_reported_improvement", "change_score_type_mean" = "change_score_type_sd"))

# Modify axis labels by creating a named vector
improvement_levels <- c("Much Worse" = 0, "Little Worse" = 1, "No Change" = 2, "Little Better" = 3, "Much Better" = 4)

# Ensure the x-axis variable is a factor with levels
long_bar_data_step2 <- long_bar_data_step2 %>%
  mutate(retrospectively_reported_improvement = factor(retrospectively_reported_improvement, 
                                                        levels = improvement_levels))

# Plotting
long_bar_data_step2 %>%
  ggplot(aes(x = retrospectively_reported_improvement, 
             y = mean, 
             fill = change_score_type_mean)) +
  geom_col(position = position_dodge(width = 0.8), width = 0.7) +
  geom_errorbar(aes(ymin = mean - sd, 
                    ymax = mean + sd), 
                position = position_dodge(width = 0.8), 
                width = 0.2, 
                size = 0.5, 
                color = "darkgrey") +
  scale_x_discrete(labels = names(improvement_levels)) +  # Use named vector for labels
  scale_fill_manual(values = c("gad7" = "#117733", 
                               "phq9" = "#88CCEE", 
                               "wsas" = "#882255")) +
  theme_minimal() +
  labs(title = "Retrospectively Reported Improvement by Different Change Scores", 
       x = "Retrospective Improvement",
       y = "Mean Change in Clinical Questionnaire Scores", 
       fill = "Clinical Questionnaire Name") +
  theme(panel.spacing = unit(1, "points"),
        panel.border = element_blank(),
        axis.line = element_line(),
        legend.text = element_text(size = 12),
        axis.text.x = element_text(size = 10),
        axis.text.y = element_text(size = 12),
        strip.background = element_blank(),
        strip.placement = "outside",
        legend.position = "right",
        plot.title = element_text(hjust = 0.5))

```


```{r}
##Computing the correlations between helpfulness and improvement

#Converting the helpfulness column to a numeric value
class(port_preprocessed_data$retrospectively_treatment_helpful)

port_preprocessed_data$retrospectively_treatment_helpful <- as.numeric(as.character(port_preprocessed_data$retrospectively_treatment_helpful))

corr_improvement_helpfulness <- cor.test(port_preprocessed_data$retrospectively_reported_improvement, port_preprocessed_data$retrospectively_treatment_helpful, method = "pearson")

print(corr_improvement_helpfulness)

corr_gad7_phq9_first <- cor.test(port_preprocessed_data$first_gad7, port_preprocessed_data$first_phq9, method = "pearson")

corr_gad7_phq9_last <- cor.test(port_preprocessed_data$last_treatment_gad7, port_preprocessed_data$last_treatment_phq9, method = "pearson")

print(corr_gad7_phq9_last)

```