mockstudy_analysis.Rmd

---
title: "Chisq and basic graphs from mockstudy"
author: "Thomas Mock, adapted from Peter Higgins"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, fig.retina = 2)
library(survival)
library(survminer)
library(patchwork)
library(tidyverse)
library(janitor)
library(gt)
library(gtsummary)
mockstudy <- read_csv(here::here("data", "mockdata.csv"))
```

### Take a look at the data

```{r glimpse}
glimpse(mockstudy)
```

## Set up a chisquared table

### By hand, with `janitor` + R

```{r chisq, messages=FALSE}
results <- mockstudy %>% 
  tabyl(arm, fu_stat) %>% 
  column_to_rownames('arm') %>% 
  chisq.test()
```

We can check some of the results

```{r}
results$statistic
results$parameter
results$p.value
```

We can also convert this to a basic table.

```{r table, warning=FALSE}
outcome_table <- mockstudy %>% 
  tabyl(arm, fu_stat)

names(outcome_table) <- c("Study Arm", "Lived", "Died")
outcome_table %>% 
  knitr::kable()
```

### More detailed table with `gtsummary`

```{r}
mock_tbl <- mockstudy %>% 
  select(arm, age, sex, bmi, fu_stat)

chi_tbl <- mock_tbl %>% 
  tbl_summary(by = arm) %>% 
  add_p()

chi_tbl
```

## Study Results

This is a statement of study results. <br>

### Inline text

In the evaluation of followup status by study arm, the null hypothesis of independence was rejected, with a chi-squared statistic of `r round(results$statistic,2)`, with `r results$parameter` degrees of freedom, and a p value of `r results$p.value`, using the `r results$method` method.


### Start with a barplot
for percent survival
tag it as panel A for a multipanel plot

```{r survival_barplot}
mockstudy %>% 
  group_by(arm) %>% 
  summarize(surv = length(which(fu_stat==1)),
         died = length(which(fu_stat==2)),
         pct_surv = surv*100/(died+surv)) %>% 
  select(arm, surv, died, pct_surv) %>% 
  ggplot() +
  aes(x=arm, y = pct_surv, fill=arm) +
  geom_bar(stat= 'identity') +
  labs(y= "Percent Survived", x= "Study Arm", tag ="A") +
  theme_minimal() +
  scale_fill_manual(values = c("black", "blue", "grey80")) ->
p1
```

### Now add a boxplot, make it multipanel
tagged as panel B

```{r survivaltime_boxplot}
mockstudy %>% 
  group_by(arm) %>% 
ggplot() +
  aes(x=arm, y = fu_time, fill=arm) +
  geom_jitter(width =0.25, alpha=0.5) +
  geom_violin(alpha =0.3) +
  labs(y= "Survival Time in \nDays (Censored)", x= "Study Arm", tag = "B") +
  theme_minimal() +
  scale_fill_manual(values = c("black", "blue", "grey80")) ->
p2
p1 + p2 + plot_layout(ncol=1, heights = c(4,4))
```


### Now add a survival curve
For some reason, patchwork does not work with this survival curve

```{r survival_curves}
surv_fit <- survfit(formula = Surv(fu_time, fu_stat) ~ arm, data= mockstudy) 
ggsurvplot(surv_fit,
           pval = TRUE, conf.int = TRUE,
           risk_table = TRUE,
           risk_table_col = "strata",
           linetype = "strata",
           surv.median.line = "hv",
           ggtheme = theme_bw(),
           palette = c("red", "blue", "green")) 
```