Distributions

Often there is information to be found in the way cases differ one from the other.

A distribution is a representation of the patterns of spread of the cases.

One variable

Categorical: how many cases at each level?

Order: help the eye

Quantitative

No discrete levels. The fundamental display is density

Show some densities.

Two or more variables: Displaying relationships

A distribution template

Aesthetics: 1, 2, or three variables

• x = The first variable: always!
• y = The count or density or another variable
• group = or fill = or color = Still another variable

Positions

• Dodged (the default)
• Stacked
• Proportion

Facets

One or two more categorical variables

• facet_wrap( ~ var)
• facet_grid(var1 ~ var2)

Variable types

interval —- ordered —– categorical

1. Quantitative: a natural order and interval
2. Ordinal: a natural order but not an interval
3. Categorical: no natural order

Order

The different levels of the variable have a natural order; it’s natural to say that one level is higher than another or that a level is in-between two other levels.

• Temperature
• Age
• Likert scale (strongly disagree, disagree, neutral, …)
• Income bracket, e.g. 0-25K, 25K-40K, 40K-75K, 75K-150K
• Numerical scale, e.g. pain level of 1 to 10
• but not, e.g. Language, Nationality, …

Interval

For variables with a natural order, does the interval between levels have a specific meaning?

• Temperature: Yes
• Age: Yes
• Likert scale: not really, but sometimes assumed
• Income bracket: No
• Language: No! (It’s not even ordered!)

NHANES

NHANES package has the NHANES data table

Categorical variables:

• Gender: no order (therefore not interval)
• Education: ordered but not interval
• SexOrientation: not really ordered (therefore not interval)
• Work: not really ordered (therefore not interval)
• MaritalStatus: no order (therefore not interval)
• HHIncome: ordered but not interval

Quantitative variables

• UrineVol1 ordered and interval
• Age: ordered and interval

Single variables

Interval (hence ordered) — use density

## Warning: Removed 987 rows containing non-finite values (stat_density).

Not interval — use counts

But respect the order if there is one.

1. Quantitative: a natural order and interval
2. Ordinal: a natural order but not an interval
3. Categorical: no natural order

Two variables

How could you make this better?

## Warning: Removed 987 rows containing non-finite values (stat_density).

## Warning: Removed 987 rows containing non-finite values (stat_density).

## Warning: Removed 987 rows containing non-finite values (stat_density).

## Warning: Width not defined. Set with `position_dodge(width = ?)`

## Warning: Removed 987 rows containing non-finite values (stat_density2d).

## Warning: Removed 987 rows containing non-finite values (stat_binhex).

## Warning: Removed 987 rows containing non-finite values (stat_smooth).