ExerciseRelationsBar_training-studies.Rmd

---
output:
  github_document:
    toc: true
    toc_depth: 1
    html_preview: false

title: "R Notebook for visualizing exercise effects on learning and memory"
---

# Setup

```{r, message=FALSE, warning=FALSE}

rm(list=ls(all=TRUE))  #clear previous
library(dplyr)
library(tidyr)
library(ggplot2)
library(RColorBrewer)


interventions = read.csv("InterventionOutcomesR.csv", header = TRUE, sep = ",")
interventions <-subset(interventions,interventions$MetaInclude == "Y")

# keep monti for relational memory graph
interventions_monti = read.csv("InterventionOutcomesR.csv", header = TRUE, sep = ",")
interventions_monti <- subset(interventions_monti,interventions_monti$StudyName == "Monti2012")
interventions <- rbind(interventions,interventions_monti)

# keep only intervention design as independent variable, keep only intervention for comparison and not many of others
interventions <-subset(interventions,interventions$IndependentVariable == "AerobicTraining")

# exclude Major Depressive Disorder
interventions <-subset(interventions,interventions$CognitiveStatus != "MajorDepressiveDisorder")

# keep only intervention outcomes that are cognitive/memory, exclude verbal fluency
interventions <-subset(interventions, interventions$DependentVariable == "VerbalPairedAssociates" | interventions$DependentVariable == "PairedAssociates" |interventions$DependentVariable == "LogicalMemory" | interventions$DependentVariable == "ComplexFigure" | interventions$DependentVariable == "ListLearning" | interventions$DependentVariable == "RAVLT" | interventions$DependentVariable == "CVLT" | interventions$DependentVariable == "AVLT" | interventions$DependentVariable == "HVLT" | interventions$DependentVariable == "Relational")


# keep only delayed recall
interventions <-subset(interventions,interventions$DependentType != "Immediate")

# remove linktype NA or unknown age
interventions <-subset(interventions,interventions$LinkType != "NA")
interventions <-subset(interventions,interventions$AgeGroup != "NA")

```



# Tasks

```{r}
# create a task variable to collapse across studies

interventions$DependentConstruct <- interventions$DependentVariable

interventions$DependentConstruct <- ifelse(interventions$DependentConstruct == "AVLT","WordListRecall",
                                    ifelse(interventions$DependentConstruct == "CVLT","WordListRecall",
                                    ifelse(interventions$DependentConstruct == "HVLT","WordListRecall",
                                    ifelse(interventions$DependentConstruct == "ListLearning","WordListRecall",
                                    ifelse(interventions$DependentConstruct == "RAVLT","WordListRecall",
                                    ifelse(interventions$DependentConstruct == "LogicalMemory","StoryRecall",
                                    ifelse(interventions$DependentConstruct == "Relational","Relational",
                                    ifelse(interventions$DependentConstruct == "VerbalPairedAssociates","Relational",
                                    ifelse(interventions$DependentConstruct == "PairedAssociates","Relational",
                                    ifelse(interventions$DependentConstruct == "ComplexFigure","Visuospatial","Other"))))))))))

interventions$CognitiveStatus<-factor(interventions$CognitiveStatus,levels=c("CognitivelyNormal","MCI","Dementia"),labels=c("CogNormal","MCI","Dementia"))


```



# Tally and plot intervention outcomes

```{r}

intervention_plot <- interventions  %>% select(StudyName,IndependentVariable,IndependentType,DependentConstruct,DependentVariable,DependentType,DependentSubType,LinkType,TotalSampleSize,AgeGroup,CognitiveStatus)

```



```{r}

data_in_fig1a <- intervention_plot %>% filter(CognitiveStatus=="CogNormal") %>%
        filter(IndependentVariable=="AerobicTraining") %>%
        filter(AgeGroup=="Older" | AgeGroup=="MiddleAge") 

write.csv(data_in_fig1a,"data_in_fig1a.csv", row.names=FALSE, na="")

```


```{r}
data_in_fig1a$LinkType <-as.factor(data_in_fig1a$LinkType)
data_in_fig1a$DependentConstruct <-as.factor(data_in_fig1a$DependentConstruct)
data_in_fig1a$DependentConstruct <- factor(data_in_fig1a$DependentConstruct, levels=c("Visuospatial","WordListRecall","StoryRecall","Relational"))
levels(data_in_fig1a$DependentConstruct)

levels(data_in_fig1a$LinkType)
data_in_fig1a$LinkType <- droplevels(data_in_fig1a$LinkType)
levels(data_in_fig1a$LinkType)
```



```{r}
# cross construct by linktype
cross_construct_table <- table(data_in_fig1a$LinkType,data_in_fig1a$DependentConstruct)
addmargins(cross_construct_table)
proportions <- round(100*prop.table(cross_construct_table,2),digits=0)
proportions
```

# Number of participants

```{r}
data_in_fig1a$TotalSampleSize <- as.numeric(data_in_fig1a$TotalSampleSize)

sample_size <- data_in_fig1a %>%
        group_by(DependentConstruct) %>%
        summarise(TotalSampleSize = sum(TotalSampleSize))

sample_size
```


# Plot data

```{r}
proportions.df = as.data.frame(proportions)

colnames(proportions.df)[colnames(proportions.df)=="Var1"] <- "Result"
colnames(proportions.df)[colnames(proportions.df)=="Var2"] <- "TaskConstruct"

ggplot(proportions.df, aes(x=TaskConstruct, y=Freq, fill = Result)) +
        geom_bar(stat="identity", color="black") +
        labs(y = NULL, fill = NULL, 
         title = "Percentage of training studies with positive outcome by task construct") +
        scale_fill_brewer(palette="Oranges") +
        guides(fill = guide_legend(reverse=TRUE)) +
        theme_classic() +
        theme(axis.line = element_blank(),
          axis.ticks = element_blank(),
          plot.title = element_text(hjust = 0.5, color = "#666666"))


# order the constructs with relevel in the order of lowest to highest percentage of positives:
# visuospatial, Word list recall, Story Recall, Relational
# add % and n of studies and participants onto the graph manually for best placement

ggsave(filename="LiteratureMemoryOutcomes_stacked-bar.pdf",width=8,units=c("in"),dpi=900)
```