This lesson is still being designed and assembled (Pre-Alpha version)

Starting with data

Overview

Teaching: min
Exercises: min
Questions
Objectives
  • Load external data from a .csv file into a data frame.

  • Describe what a data frame is.

  • Summarize the contents of a data frame.

  • Use indexing to subset specific portions of data frames.

  • Describe what a factor is.

  • Convert between strings and factors.

  • Reorder and rename factors.

  • Change how character strings are handled in a data frame.

  • Format dates.

Presentation of the ADNI Dataset

The data we will be using today is from the Alzheimer’s Disease Neuroimaging Initiative (ADNI).

ADNI is a global research study that actively supports the investigation and development of treatments that slow or stop the progression of AD. In this multisite longitudinal study, researchers at 63 sites in the US and Canada track the progression of AD in the human brain with clinical, imaging, genetic and biospecimen biomarkers through the process of normal aging, early mild cognitive impairment (EMCI), and late mild cognitive impairment (LMCI) to dementia or AD. The overall goal of ADNI is to validate biomarkers for use in Alzheimer’s disease clinical treatment trials. The dataset is stored as a comma separated value (CSV) file.

Each row holds information for a single patient visit, and the columns represent patient attributes.

Column Description
PTID unique patient identifier
AGE_mod a modified patient age for deidentification
PTGENDER the patient’s gender
PTMARRY the patient’s marital status
PTEDUCATE the patient’s educational status
DX the patient’s diagnosis
SITE the study site the patient was seen at
EXAMDATE_mod the date the patient was seen
WholeBrain the volume of the patient’s brain at a given visit
WholeBrain_bl the volume of the patient’s brain at the baseline visit
Hippocampus the volume of the patient’s hippocampus at a given visit
APOE4 the number of risk alleles the pateient has in the APOE4 gene
TAU_mod Level of the TAU biomarker in cerebrospinal fluid

For a full data inventory, see (http://adni.loni.usc.edu/data-samples/adni-data-inventory/)

We will use read.csv() to load into memory the content of the CSV file as an object of class data.frame.

You should have downloaded the data for this lesson at the beginning of the day. It should be in a folder called “data”.

You are now ready to load the data:

adni_r <- read.csv("data/ADNIMERGE.csv")

This statement doesn’t produce any output because, as you might recall, assignments don’t display anything. If we want to check that our data has been loaded, we can see the contents of the data frame by typing its name: adni_r.

Wow… that was a lot of output. At least it means the data loaded properly. Let’s check the top (the first 6 lines) of this data frame using the function head():

head(adni_r)
        PTID VISCODE AGE_mod PTGENDER PTMARRY PTEDUCAT       DX SITE
1 011_S_0002      bl    74.3     Male Married       16       CN   11
2 011_S_0003      bl    81.3     Male Married       18 Dementia   11
3 011_S_0003     m06    81.3     Male Married       18 Dementia   11
4 011_S_0003     m12    81.3     Male Married       18 Dementia   11
5 011_S_0003     m24    81.3     Male Married       18 Dementia   11
6 022_S_0004      bl    67.5     Male Married       10      MCI   22
  EXAMDATE_mod WholeBrain WholeBrain_bl Hippocampus APOE4 TAU_mod
1   2005-08-18    1229740       1229740        8336     0    -4.0
2   2005-10-11    1129830       1129830        5319     1   239.7
3   2006-03-11    1100060       1129830        5446     1    -4.0
4   2006-09-29    1095640       1129830        5157     1   251.7
5   2007-09-06    1088560       1129830        5139     1    -4.0
6   2005-11-10    1154980       1154980        6869     0   153.1
## Try also
View(adni_r)

Note

read.csv assumes that fields are delineated by commas, however, in several countries, the comma is used as a decimal separator and the semicolon (;) is used as a field delineator. If you want to read in this type of files in R, you can use the read.csv2 function. It behaves exactly like read.csv but uses different parameters for the decimal and the field separators. If you are working with another format, they can be both specified by the user. Check out the help for read.csv() by typing ?read.csv to learn more. There is also the read.delim() for in tab separated data files. It is important to note that all of these functions are actually wrapper functions for the main read.table() function with different arguments. As such, the ANDI data above could have also been loaded by using read.table() with the separation argument as ,. The code is as follows: adni_r <- read.table(file="data/ADNIMERGE.csv", sep=",", header=TRUE). The header argument has to be set to TRUE to be able to read the headers as by default read.table() has the header argument set to FALSE.

What are data frames?

Data frames are the de facto data structure for most tabular data, and what we use for statistics and plotting.

A data frame can be created by hand, but most commonly they are generated by the functions read.csv() or read.table(); in other words, when importing spreadsheets from your hard drive (or the web).

A data frame is the representation of data in the format of a table where the columns are vectors that all have the same length. Because columns are vectors, each column must contain a single type of data (e.g., characters, integers, factors). For example, here is a figure depicting a data frame comprising a numeric, a character, and a logical vector.

We can see this when inspecting the structure of a data frame with the function str():

str(adni_r)

Inspecting data.frame Objects

We already saw how the functions head() and str() can be useful to check the content and the structure of a data frame. Here is a non-exhaustive list of functions to get a sense of the content/structure of the data. Let’s try them out!

Note: most of these functions are “generic”, they can be used on other types of objects besides data.frame.

Challenge

Based on the output of str(adni_r), can you answer the following questions?

  • What is the class of the object adni_r?
  • How many rows and how many columns are in this object?
  • How many types of diagnoses have been observed during this study?
str(adni_r)
'data.frame':	13900 obs. of  14 variables:
 $ PTID         : Factor w/ 2152 levels "002_S_0295","002_S_0413",..: 241 242 242 242 242 531 531 531 531 531 ...
 $ VISCODE      : Factor w/ 26 levels "bl","m03","m06",..: 1 1 3 7 14 1 3 7 13 16 ...
 $ AGE_mod      : num  74.3 81.3 81.3 81.3 81.3 67.5 67.5 67.5 67.5 67.5 ...
 $ PTGENDER     : Factor w/ 2 levels "Female","Male": 2 2 2 2 2 2 2 2 2 2 ...
 $ PTMARRY      : Factor w/ 5 levels "Divorced","Married",..: 2 2 2 2 2 2 2 2 2 2 ...
 $ PTEDUCAT     : int  16 18 18 18 18 10 10 10 10 10 ...
 $ DX           : Factor w/ 4 levels "-4","CN","Dementia",..: 2 3 3 3 3 4 4 4 4 4 ...
 $ SITE         : int  11 11 11 11 11 22 22 22 22 22 ...
 $ EXAMDATE_mod : Factor w/ 4213 levels "2005-08-18","2005-09-03",..: 1 4 85 258 595 13 108 284 499 1006 ...
 $ WholeBrain   : int  1229740 1129830 1100060 1095640 1088560 1154980 1116280 1117390 1095210 1085350 ...
 $ WholeBrain_bl: int  1229740 1129830 1129830 1129830 1129830 1154980 1154980 1154980 1154980 1154980 ...
 $ Hippocampus  : int  8336 5319 5446 5157 5139 6869 6439 6451 6373 6213 ...
 $ APOE4        : int  0 1 1 1 1 0 0 0 0 0 ...
 $ TAU_mod      : num  -4 240 -4 252 -4 ...

Solution

  • class: data frame
  • how many rows: 13,900, how many columns: 112
  • how many types of diagnoses (DX): 4

Indexing and subsetting data frames

Our ADNI data frame has rows and columns (it has 2 dimensions), if we want to extract some specific data from it, we need to specify the “coordinates” we want from it. Row numbers come first, followed by column numbers. However, note that different ways of specifying these coordinates lead to results with different classes.

# first element in the first column of the data frame (as a vector)

adni_r[1, 1]
# first element in the 6th column (as a vector)
adni_r[1, 6]
# first column of the data frame (as a vector)
adni_r[, 1]
# first column of the data frame (as a data.frame)
adni_r[1]
# first three elements in the 7th column (as a vector)
adni_r[1:3, 7]
# the 3rd row of the data frame (as a data.frame)
adni_r[3, ]
# equivalent to head_adni <- head(adni_r)
head_adni <- adni_r[1:6, ]

: is a special function that creates numeric vectors of integers in increasing or decreasing order, test 1:10 and 10:1 for instance.

You can also exclude certain indices of a data frame using the “-” sign:

adni_r[, -1]          # The whole data frame, except the first column
adni_r[-c(7:34786), ] # Equivalent to head(adni_r)

Data frames can be subset by calling indices (as shown previously), but also by calling their column names directly:

adni_r["DX"]       # Result is a data.frame
adni_r[, "DX"]     # Result is a vector
adni_r[["DX"]]     # Result is a vector
adni_r$DX          # Result is a vector

In RStudio, you can use the autocompletion feature to get the full and correct names of the columns.

Challenge

  1. Create a data.frame (adni_200) containing only the data in row 200 of the adni_r dataset.

  2. Notice how nrow() gave you the number of rows in a data.frame?

    • Use that number to pull out just that last row in the data frame.
    • Compare that with what you see as the last row using tail() to make sure it’s meeting expectations.
    • Pull out that last row using nrow() instead of the row number.
    • Create a new data frame (adni_last) from that last row.
  3. Use nrow() to extract the row that is in the middle of the data frame. Store the content of this row in an object named adni_middle.

  4. Combine nrow() with the - notation above to reproduce the behavior of head(adni_r), keeping just the first through 6th rows of the adni dataset.

Solution

# 1.
adni <- adni[200, ]
Error in eval(expr, envir, enclos): object 'adni' not found
# 2.
# Saving `n_rows` to improve readability and reduce duplication
n_rows <- nrow(adni)
Error in nrow(adni): object 'adni' not found
adni_last <- adni[n_rows, ]
Error in eval(expr, envir, enclos): object 'adni' not found
# 3.
adni_middle <- adni[n_rows / 2, ]
Error in eval(expr, envir, enclos): object 'adni' not found
# 4.
adni_head <- adni[-(7:n_rows), ]
Error in eval(expr, envir, enclos): object 'adni' not found

Factors

When we did str(adni_r) we saw that several of the columns consist of integers. The columns PTID, PTMARRY, DX, PTGENDER, … however, are of a special class called factor. Factors are very useful and actually contribute to making R particularly well suited to working with data. So we are going to spend a little time introducing them.

Factors represent categorical data. They are stored as integers associated with labels and they can be ordered or unordered. While factors look (and often behave) like character vectors, they are actually treated as integer vectors by R. So you need to be very careful when treating them as strings.

Once created, factors can only contain a pre-defined set of values, known as levels. By default, R always sorts levels in alphabetical order. For instance, if you have a factor with 2 levels:

gender <- factor(c("Male", "Female", "Female", "Male"))

R will assign 1 to the level "Female" and 2 to the level "Male" (because f comes before m, even though the first element in this vector is "Male"). You can see this by using the function levels() and you can find the number of levels using nlevels():

levels(gender)
nlevels(gender)

Sometimes, the order of the factors does not matter, other times you might want to specify the order because it is meaningful (e.g., “low”, “medium”, “high”), it improves your visualization, or it is required by a particular type of analysis. Here, one way to reorder our levels in the gender vector would be:

gender # current order
[1] Male   Female Female Male  
Levels: Female Male
gender <- factor(gender, levels = c("Male", "Female"))
gender # after re-ordering
[1] Male   Female Female Male  
Levels: Male Female

In R’s memory, these factors are represented by integers (1, 2, 3), but are more informative than integers because factors are self describing: "Female", "Male" is more descriptive than 1, 2. Which one is “Male”? You wouldn’t be able to tell just from the integer data. Factors, on the other hand, have this information built in. It is particularly helpful when there are many levels (like the species names in our example dataset).

Converting factors

If you need to convert a factor to a character vector, you use as.character(x).

as.character(gender)

In some cases, you may have to convert factors where the levels appear as numbers (such as concentration levels or years) to a numeric vector. For instance, in one part of your analysis the years might need to be encoded as factors (e.g., comparing average ages between genders) but in another part of your analysis they may need to be stored as numeric values (e.g., doing math operations on the years). This conversion from factor to numeric is a little trickier. The as.numeric() function returns the index values of the factor, not its levels, so it will result in an entirely new (and unwanted in this case) set of numbers. One method to avoid this is to convert factors to characters, and then to numbers.

Another method is to use the levels() function. Compare:

year_fct <- factor(c(1990, 1983, 1977, 1998, 1990))
as.numeric(year_fct)               # Wrong! And there is no warning...
as.numeric(as.character(year_fct)) # Works...
as.numeric(levels(year_fct))[year_fct]    # The recommended way.

Notice that in the levels() approach, three important steps occur:

Renaming factors

When your data is stored as a factor, you can use the plot() function to get a quick glance at the number of observations represented by each factor level. Let’s look at the number of males and females captured over the course of the experiment:

## bar plot of the number of men and women studied:
plot(adni_r$PTGENDER)

plot of chunk 02-rename

You may need to rename the levels of your factors. Here’s an example of how to rename the “Female” factor to “Female”

gender <- adni_r$PTGENDER
head(gender)
[1] Male Male Male Male Male Male
Levels: Female Male
levels(gender)
[1] "Female" "Male"  
levels(gender)[1] <- "female"
levels(gender)
[1] "female" "Male"  
head(gender)
[1] Male Male Male Male Male Male
Levels: female Male

Challenge

  • Rename “female” and “male” to “F” and “M” respectively.
  • Can you recreate the barplot such that “M” is before “F”?

Solution

levels(gender) <- c("F", "M")
gender <- factor(gender, levels = c( "M", "F"))
plot(gender)

plot of chunk unnamed-chunk-3

Using stringsAsFactors=FALSE

By default, when building or importing a data frame, the columns that contain characters (i.e. text) are coerced (= converted) into factors. Depending on what you want to do with the data, you may want to keep these columns as character. To do so, read.csv() and read.table() have an argument called stringsAsFactors which can be set to FALSE.

In most cases, it is preferable to set stringsAsFactors = FALSE when importing data and to convert as a factor only the columns that require this data type.

## Compare the difference between our data read as `factor` vs `character`.
adni_r <- read.csv("data/ADNIMERGE.csv", stringsAsFactors = TRUE)
str(adni_r)
adni_r <- read.csv("data/ADNIMERGE.csv", stringsAsFactors = FALSE)
str(adni_r)
## Convert the column "plot_type" into a factor
adni_r$SITE <- factor(adni_r$SITE)

Challenge

  1. We have seen how data frames are created when using read.csv(), but they can also be created by hand with the data.frame() function. There are a few mistakes in this hand-crafted data.frame. Can you spot and fix them? Don’t hesitate to experiment!

     animal_data <- data.frame(
               animal = c(dog, cat, sea cucumber, sea urchin),
               feel = c("furry", "squishy", "spiny"),
               weight = c(45, 8 1.1, 0.8)
               )
    
  2. Can you predict the class for each of the columns in the following example? Check your guesses using str(country_climate):

    • Are they what you expected? Why? Why not?
    • What would have been different if we had added stringsAsFactors = FALSE when creating the data frame?
    • What would you need to change to ensure that each column had the accurate data type?
     country_climate <- data.frame(
            country = c("Canada", "Panama", "South Africa", "Australia"),
            climate = c("cold", "hot", "temperate", "hot/temperate"),
            temperature = c(10, 30, 18, "15"),
            northern_hemisphere = c(TRUE, TRUE, FALSE, "FALSE"),
            has_kangaroo = c(FALSE, FALSE, FALSE, 1)
            )
    

Solution

1.

  • missing quotations around the names of the animals
  • missing one entry in the feel column (probably for one of the furry animals)
  • missing one comma in the weight column

2.

  • country, climate, temperature, and northern_hemisphere are factors; has_kangaroo is numeric
  • using stringsAsFactors = FALSE would have made character vectors instead of factors
  • removing the quotes in temperature and northern_hemisphere and replacing 1 by TRUE in the has_kangaroo column would give what was probably intended

The automatic conversion of data type is sometimes a blessing, sometimes an annoyance. Be aware that it exists, learn the rules, and double check that data you import in R are of the correct type within your data frame. If not, use it to your advantage to detect mistakes that might have been introduced during data entry (for instance, a letter in a column that should only contain numbers).

Learn more in this RStudio tutorial

Formatting Dates

One of the most common issues that new (and experienced!) R users have is converting date and time information into a variable that is appropriate and usable during analyses. As a reminder from earlier in this lesson, the best practice for dealing with date data is to ensure that each component of your date is stored as a separate variable. Using str(), We can confirm that our data frame has a separate column for day, month, and year, and that each contains integer values.

str(adni_r)

We are going to use the ymd() function from the package lubridate (which belongs to the tidyverse; learn more here). . lubridate gets installed as part as the tidyverse installation. When you load the tidyverse (library(tidyverse)), the core packages (the packages used in most data analyses) get loaded. lubridate however does not belong to the core tidyverse, so you have to load it explicitly with library(lubridate)

Start by loading the required package:

library(lubridate)

ymd() takes a vector representing year, month, and day, and converts it to a Date vector. Date is a class of data recognized by R as being a date and can be manipulated as such. The argument that the function requires is flexible, but, as a best practice, is a character vector formatted as “YYYY-MM-DD”.

Let’s create a date object and inspect the structure:

my_date <- ymd("2015-01-01")
str(my_date)

Now let’s paste the year, month, and day separately - we get the same result:

# sep indicates the character to use to separate each component
my_date <- ymd(paste("2015", "1", "1", sep = "-"))
str(my_date)

Now we apply this function to the ADNI dataset. Change the EXAMDATE_mod column to character:

adni_r$EXAMDATE_mod<-as.character(adni_r$EXAMDATE_mod)

This character vector can be used as the argument for ymd():

adni_r$EXAMDATE_mod<-ymd(adni_r$EXAMDATE_mod)

Let’s make sure everything worked correctly. One way to inspect the new column is to use summary():

summary(adni_r$EXAMDATE_mod)
        Min.      1st Qu.       Median         Mean      3rd Qu. 
"2005-08-18" "2008-10-15" "2012-04-06" "2011-11-07" "2013-11-18" 
        Max. 
"2019-04-25" 

Key Points