Messy data? Get used to it

• Since lab 1, the data we’ve been using has been pretty clean.
• Why do we call it clean?
  • Variables were named so we could understand what they were about.
  • There didn’t seem to be any typos in the values.
  • Numerical variables were considered numbers.
  • Categorical variables were composed of categories.
• Unfortunately, more often than not, data is messy until YOU clean it.
• In this lab, we’ll learn a few essentials for cleaning dirty data.

Messy data?

• What do we mean by messy data?
• Variables might have non-descriptive names
  • Var01, V2, a, …
• Categorical variables might have misspelled categories
  • “blue”, “Blue”, “blu”, …
• Numerical variables might have been input incorrectly. For example, if we’re talking about people’s height in inches:
  • 64.7, 6.86, 676, …
• Numerical variables might be incorrectly coded as categorical variables (or vice-versa)
  • “64.7”, “68.6”, “67.6”

The American Time Use Survey

• To show you what dirty data looks like, we’ll check out the American Time Use Survey, or ATU survey.
• What is ATU survey?
  • It’s a survey conducted by the US government (specifically the Bureau of Labor Statistics).
  • They survey thousands of people to find out exactly what activities they do throughout a single day.
  • These thousands of people combined together give an idea about how much time the typical person living in the US spends doing various activities.

Load and go:

• Type the following commands into your console:

data(atu_dirty)

View(atu_dirty)

• Just by viewing the data, what parts of our ATU data do you think need cleaning?

Description of ATU Variables

• The description of the actual variables:
  • caseid: Anonymous ID of survey taker.
  • V1: The age of the respondent.
  • V2: The sex of the respondent.
  • V3: Whether the person is employed full-time or part-time.
  • V4: Whether the person has a physical difficulty.
  • V5: How long the person sleeps, in minutes.
  • V6: How long the survey taker spent on homework, in minutes.
  • V7: How long the respondent spent socializing, in minutes.

New name, same old data

• To fix the variable names, we need to assign a new set of names in place of the old ones.
  • Below is an example of the rename function:

atu_cleaner <- rename(atu_dirty, age = V1,
                       sex = V2)

• Use the example code and the variable information on the previous slide to rename the rest of the variables in atu_dirty.
• Write down the new names you chose for the rest of the variables in atu_dirty.
  • Names should be short, contain no spaces and describe what the variable is related to. So use abbreviations to your heart’s content.

Next up: Strings

• In programming, a string is sort of like a word.
  • It’s a value made up of characters (i.e. letters)
• The following are examples of strings. Notice that each string has quotes before and after.

"string"

"A1B2c3"

"Hot Cocoa"

"0015"

Numbers are words? (Sometimes)

• In some cases, R will treat values that look like numbers as if they were strings.
• Sometimes we do this on purpose.
  • For example, we can code Yes/No variables as "1"/"0".
• Sometimes we don’t mean for this to happen.
  • The number of siblings a person has should not be a string.
• Look at the structure of your data and the variable descriptions from a few slides back:
  • Write down the variables that should be numeric but are improperly coded as strings or characters.

Changing strings into numbers

• To fix this problem, we need to tell R to think of our “numeric” variables as numeric variables.
• We can do this with the as.numeric function.
  • An example using this function is below:

as.numeric("3.14")

## [1] 3.14

• Notice: We started with a string, "3.14", but as.numeric was able to turn it back into a number.

Mutating in action

• Look at the variables you thought should be numeric and select one. Then fill in the blanks below to see how we can correctly code it as a number:

atu_cleaner <- mutate(atu_cleaner, 
                 age = as.numeric(age),
                 ___ = as.numeric(___))

• Once you have this code working, use a similar line of code to correctly code the other numeric variables as numbers.

Deciphering Categorical Variables

• We mentioned earlier that we sometimes code categorical variables as numbers.
  • For example, our sex variable uses "01" and "02" for "Male" and "Female", respectively.
• It’s often much easier to analyze and interpret when we use more descriptive categories, such as "Male" and "Female".

Factors and Levels

• R has a special name for categorical variables, called factors.
• R also has a special name for the different categories of a categorical variable.
  • The individual categories are called levels.
• To see the levels of sex and their counts, type:

tally(~sex, data = atu_cleaner)

• Use similar code as we used above to write down the levels for the three factors in our data.

A level by any other name…

• If we know that ‘01’ means ‘Male’ and ‘02’ means ‘Female’ then we can use the following code to recode the levels of sex.
• Type the following command into your console:

atu_cleaner <- mutate(atu_cleaner, 
                      sex = recode(sex, 
                         "01" = "Male", 
                         "02" = "Female"))

• This code is definitely a bit of a mouthful. Let’s break it down.

Allow me to explain

atu_cleaner <- mutate(atu_cleaner, 
                      sex = recode(sex, 
                         "01" = "Male", 
                         "02" = "Female"))

• This code is saying:
  • Replace my current version of atu_cleaner…
  • with a mutated one where …
  • the sex variable’s levels …
  • have been recoded…”
  • where "01" will now be "Male"…
  • and "02" will now be "Female".

Finish it off!

• Recode the categorical variable about whether the person surveyed had a physical challenge or not. The coding is currently:

  • "01": Person surveyed did not have a physical challenge.
  • "02": Person surveyed did have a physical challenge.

• Write a script that:

  1. Loads the atu_dirty dataset
  2. Cleans the data as we have in this lab
  3. Saves a copy of the cleaned data (see next slide).

• NOTE: You can watch this video to learn about RScripts:

The final lines

• The last few lines of your script are extremely important because they will save all of your work.

• Be sure to View your data and check its structure to make sure it looks clean and tidy before saving.

• Run the code below:

atu_clean <- atu_cleaner

• This code will create a new data frame in your Environment called atu_clean which is a final copy of atu_cleaner.

  • If atu_clean is swept from your Environment all of the changes you made will NOT be saved.
  • You would need to re-run the script to clean the data again.

• To permanently save your changes you need to save the file as an R data file or .Rda.

• Run the code below:

save(atu_clean, file = "atu_clean.Rda")

• Look in your Files pane for the atu_clean.Rda file.
  • This is as permanent copy of your clean atu data.
  • To load the data onto your Environment click on the file.
  • A pop-up window confirming the upload will appear.

Flex your skills

• Now that you have learned some cleaning data basics, it’s time to revisit the food data.

  • Import your food data onto the Environment pane.

• Run the code below:

histogram(~calories | healthy_level, data = food)

• Use the as.factor() function to convert healthy_level into a categorical variable and re-run the histogram function.

  • Notice that the healthy_level categories are now numbers as opposed to tick-marks. This is an improvement but an even better solution would be to recode the categories.

• Recode the healthy_level categories and re-run the histogram function.

  • “1” = “Very Unhealthy”
  • “2” = “Unhealthy”
  • “3” = “Neutral”
  • “4” = “Healthy”
  • “5” = “Very Healthy”

• If your food data is cleared from your Environment, the changes that you made to the healthy_level variable will not be saved.

• To save your changes permanently, save your food file as an R data file.