DATASETS.Rmd

---
title: 'Datasets and dataframes'
---

```{r, include=FALSE}
library(tidyverse)
library(tufte)
```

# (PART) Data {- #data}

# The `dataframe` {#datasets-dataframes}

A `dataframe` is a container for our data.

It's much like a spreadsheet, but with some constraints applied. 'Constraints'
might sound bad, but they're actually helpful: they make dataframes more
structured and predictable to work with. The main constraints are that:

-   Each column is a [vector](#vectors-and-lists), and so can
    [only store one type of data](#vectors-and-lists).

-   Every column has to be the same length (although missing values are
    allowed).

-   Each column must have a name.

A `tibble` is an updated version of a dataframe with a whimsical name, which is
part of the `tidyverse`. It's almost exactly the same a dataframe, but with some
rough edges smoothed off — it's safe and preferred to use `tibble` in place of
`data.frame`.

You can make a simple tibble or dataframe like this:

```{r}
data.frame(myvariable = 1:10)
```

Using a tible is much the same, but allows some extra tricks like creating one
variable from another:

```{r}
tibble(
    height_m = rnorm(10, 1.5, .2),
    weight_kg = rnorm(10, 65, 10),
    bmi = weight_kg / height_m ^ 2,
    overweight = bmi > 25
)
```

#### Using 'built in' data {- #built-in-data}

The quickest way to see a dataframe in action is to use one that is built in to
R
([this page](https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/00Index.html)
lists all the built-in datasets). For example:

```{r}
head(airquality)
```

Or

```{r}
head(mtcars)
```

In both these examples the datasets are already loaded and available to be used
with the `head()` function.

:::{.exercise}

To find a list of all the built in datasets you can type `help(datasets)` into
the console, or see
<https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/00Index.html>.

Familiarise yourself with some of the other included datasets, e.g.
`datasets::attitude`. Watch out that not all the included datasets are
_dataframes_: Some are just vectors of observations (e.g. the `airmiles` data)
and some are 'time-series', (e.g. the `co2` data)

:::

#### Looking at dataframes {- #looking-at-data}

As we've already seen, using `print(df)` within an RMarkdown document creates a
nice interactive table you can use to look at your data.

However you won't want to print your whole data file when you Knit your
RMarkdown document. The `head` function can be useful if you just want to show a
few rows:

```{r}
head(mtcars)
```

Or we can use `glimpse()` function from the `dplyr::` package (see the
[section on loading and using packages](#packages)) for a different view of the
first few rows of the `mtcars` data. This flips the dataframe so the variables
are listed in the first column of the output:

```{r}
glimpse(mtcars)
```

You can use the `pander()` function (from the `pander::` package) to format
tables nicely, for when you Knit a document to HTML, Word or PDF. For example:

```{r}
library(pander)
pander(head(airquality), caption="Tables always need a caption.")
```

See the section on [sharing and publishing](#sharing-and-publication) for more
ways to format and present tables.

Other useful functions for looking at and exploring datasets include:

-   `summary(df)`
-   `psych::describe(df)`
-   `skimr::skim(df)`

:::{.exercise}

Experiment with a few of the functions for viewing/summarising dataframes.

:::

There are also some helpful plotting functions which accept a whole dataframe as
their input:

```{r, fig.cap="Box plot of all variables in a dataset."}
boxplot(airquality)
```

```{r, fig.cap="Correlation heatmap of all variables in a dataset. Colours indicate size of the correlation between pairs of variables."}
psych::cor.plot(airquality)
```

These plots might not be worth including in a final write-up, but are very
useful when exploring your data.