fancy-reshaping.Rmd

---
title: 'Fancy reshaping'
---

```{r, include=FALSE, message=F}
library(tidyverse)
library(reshape2)
library(pander)
```

### Fancy reshaping {- #fancy-reshaping}

As noted above, it's common to combine the process of reshaping and aggregating
or summarising in the same step.

For example here we have multiple rows per person, 3 trial at time 1, and 3 more
trials at time 2:

```{r, include=F}
expt.data <- readRDS("data/expt.data.RDS") %>% filter(trial < 4)
```

```{r}
expt.data %>%
  arrange(person, time, trial) %>%
  head %>%
  pander
```

We can reshape and aggregate this in a single step using `dcast`. Here we
request the mean for each person at each time, with observations for each time
split across two columns:

```{r, message=T, warning=T}
library(reshape2)
expt.data %>%
	dcast(person~paste0('time',time),
	      fun.aggregate=mean) %>%
  head %>%
  pander
```

Here `dcast` has correctly guessed that `RT` is the value we want to aggregate
(you can specify explicitly with the `value.var=` parameter).

`dcast` knows to aggregate using the mean because we set this with the
`agg.function` parameter; this just stands for 'aggregation function'.

We don't have to request the mean though: any function will do. Here we request
the SD instead:

```{r, message=T, warning=T}
expt.data %>%
	dcast(person~time,
	      fun.aggregate=sd) %>%
  head %>%
  pander
```