Exercise 6: Subsets of data and logical operators

Logical vector and index vector

A lot of times we want to get a subset of data filtering rows or columns of a dataframe. For which we can perform logical test and get TRUE or FALSE as result. This vector of logical can then be used to subset the observations from a dataframe.

For example, Lets extract observation from bodydata with Weight greater than 80. You might be wondering why following code does not work,

isHeavy <- Weight > 80

Error in eval(expr, envir, enclos): object 'Weight' not found

But remember that, the variable Weight is a part of bodydata. We have to extract Weight from the bodydata first. In R, with and within function helps you in this respect. In the following code, with function goes inside bodydata and execute the expression Weight > 80.

isHeavy <- with(bodydata, Weight > 80)

Here the logical vector isHeavy is computed by performing a logical operation on Weight variable within bodydata. The same operation can be done as,

isHeavy <- bodydata$Weight > 80

Take a look at this variable, what is it? :

head(isHeavy)

[1] FALSE  TRUE FALSE FALSE FALSE  TRUE

Yes, it is a vector of TRUE and FALSE with same length as Weight. Here the condition has compared each element of Weight results TRUE if it is greater than 80 and FALSE if it is less than 80.

Identify the elements

We can identify which observations that are heavy by the which() function

HeavyId <- which(isHeavy)

This will return a vector of row index for the observations that are heavy, i.e. greater than 80. So how many are heavy? To find the size of a vector we can use length function.

length(HeavyId)

[1] 94

Here, 94 observations have Weight larger than 80.

Exercise

Identify who are taller than 180 and save this logical vector as an object called isTall.
Answer
```
 isTall <- with(bodydata, Height > 180) 
```
How many observations have height taller than 180?
Answer
```
 TallId <- which(isTall)
 length(TallId)
```
```
 [1] 76
```
How many observations are both tall and heavy? Here, you can use length function as above to find how many person are taller than 180.
Answer
```
 isBoth <- isHeavy * isTall
```
How is this computation done?

Here isHeavy and isTall contains TRUE and FALSE. The multiplication of logical operator results a logical vector with TRUE only if both the vectors are TRUE else FALSE.

Alternatively :
```
 isBoth <- which(isHeavy & isTall)
```
The & operator result TRUE if both isHeavy and isTall are TRUE else, FALSE which is same as previous.

Subsetting data frame

Example 1

Lets create a subset of the data called bodydataTallAndHeavy containing only the observations for tall and heavy persons as defined by isBoth.

bodydataTallAndHeavy <- bodydata[isBoth, ]

For other logical tests see help file ?Comparison

Example 2

Lets create a random subset of 50 observations. For this we first sample 50 row index randomly from all rows in bodydata. The sample function is used for the purpose. In the following code, nrow(bodydata) return the number of rows in bodydata. The sample function takes two argument x which can be a vector or a integer and size which is the size of the sample to be drawn.

idx <- sample(x = nrow(bodydata), size = 50)

Here, 50 rows are sampled from the total number of rows and the index of the selected rows are saved on vector idx.

Using this vector we can select the observations in bodydata to create a new data set called bodydataRandom as,

bodydataRandom <- bodydata[idx, ]

Here is the first five rows of bodydataRandom dataset.

head(bodydataRandom, n = 5)

    Weight Height Age Circumference
234   74.1    168  55          90.1
283   53.6    162  23          63.2
91    70.2    175  27          73.5
378   58.0    163  20          71.0
105   55.5    169  20          68.0

Exercise

Create a subset of dataset bodydata including the observation with Age larger an 55 and Circumference larger than 80. Save this dataset named subdata.

Answer

idx <- with(bodydata, Age > 55 & Circumference > 80)
subdata <- bodydata[idx, ]
subdata

    Weight Height Age Circumference
136   76.4    185  62          94.8
189   73.6    175  60          90.5
207   66.8    168  62          81.5
231   80.0    174  65          98.6

For those who are interested in playing more with data, have a look at http://r4ds.had.co.nz/transform.html