A variable is any characteristic whose value may change from one individual (or object) to another.
Data result from making observations on a single variable (called univariate data) or simultaneously on two or more variables (called multivariate data).
Note
Recall the data in Chapter 1. Each row provides information about a group of early-stage Delta smelt larvae exposed to particular light and turbidity conditions. Each row is called a case or observational unit. Here, a case is a group of larvae, rather than one larva. The columns are variables: light intensity (Light), turbidity (Turbidity), and survival percentage (Survival). For example, the first row records 56.8% survival for larvae exposed to light intensity of 4.5 \(\mu mol/m^2/s\) and turbidity of 2 NTUs.
7.2 Types of variables
Variables in a statistical study can be classified as either categorical or numerical.
A variable is categorical if the individual observations are categorical responses, such as labels, people’s attributes, opinions, etc.
A variable is numerical (or quantitative) if its corresponding observations are numerical values that are measured or counted.
Factor variables are how R stores categorical variables. The levels of a factor are its possible categories. For example, first-generation college student status might have two levels: 1 for a first-generation student and 0 for a student who is not first-generation. A categorical variable is ordinal if its categories have a natural order, such as low, medium, and high. Otherwise, it is nominal, such as county name. In these examples, R compares the light and turbidity categories as separate groups.
Note
The case-study data in Chapter 1 consist of three variables: Light, Turbidity, and Survival. - Survival is a numerical variable that records the percentage of larvae that survived. For the methods discussed here, we treat survival percentage as continuous. Continuous variables describe measurements, such as height or temperature, rather than counts. - Light and Turbidity are both factor variables, with each having levels low, medium, and high.
Figure 7.1: Breakdown of variables into their respective types
ImportantR and factor variables
When data are imported (described in Appendix C) that contains factor variables, R may not automatically recognize them as such1. If the levels of the factor variables are represented by numbers, R by default, classifies these variables as numeric. If the levels are instead letters or words, R will treat them as character variables by default.
To convert a variable to a factor variable, you can use the mutate() function in the dplyr package.
TipR functions
### mutate( data ,
### 'new variable' = 'function of variable in data frame', ...)
# data: Replace with the name of the data frame being used.
# 'new variable' = 'function of variable in data frame': Replace
# 'new variable' with name of desired variable
# and set equal to a function of
# a current variable in the data frame.
# This will add the new variable to the
# data frame. The 'new variable' name may
# also be replaced with the name of a current
# variable as this will override/replace the
# current variable.
#...: additional new variables can be added to the data frame.
#
### summary( data )
# Numeric variables: minimum, quartiles, median, mean, maximum,
# and an NA count when applicable. Factors: counts by level.
# Use sum(!is.na(x)) for the nonmissing sample size.
Note
The case study data given in Chapter 1 consist of two factor variables. The code below instructs R that Light and Turbidity are factor variables by using the function mutate(). Combined with mutate(), the function as.factor() converts a variable into a factor variable
# Import the dataset 'dssurv' from a csv file#dssurv <- read.csv("datasets/dssurv.csv")library(CSUBstats)data(dssurv)#load the dataset named `dssurv` into your current workspace# Print a summary of the dataset, including the # five-number summary for numeric variables and counts # of observations for categorical variables.summary(dssurv)#> Light Turbidity Survival #> high:8 high:8 Min. :45.02 #> low :8 low :8 1st Qu.:55.79 #> med :4 med :4 Median :59.56 #> Mean :62.22 #> 3rd Qu.:67.24 #> Max. :89.09# Load the 'dplyr' package which provides the 'mutate()' functionlibrary(dplyr)# Replace `Light` and `Turbidity` with the factor# versions. The code below tells R to convert # `Light` and `Turbidity` into factor variables.dssurv<-mutate(dssurv, Light=as.factor(Light) , Turbidity=as.factor(Turbidity))# Print a new summary of the updated dataset.summary(dssurv)#> Light Turbidity Survival #> high:8 high:8 Min. :45.02 #> low :8 low :8 1st Qu.:55.79 #> med :4 med :4 Median :59.56 #> Mean :62.22 #> 3rd Qu.:67.24 #> Max. :89.09
7.3 .csv files
To provide a consistent approach to using R, it is assumed that any research data will be saved in the tidy format using a comma separated values (.csv) file. A .csv may be viewed with a spreadsheet program such as Microsoft Excel. It is assumed readers have used an Excel spreadsheet. If not, please view the tutorial Excel Tutorial for Beginners .
To store data in a .csv file, organize your data in an Excel worksheet, such that the first row (Row 1) contains the column (variable) names and each subsequent row contains all the necessary information for each data point. Next, save the worksheet as a .csv file type.
Sample .csv file in tidy format
7.4 Tidy data
Note the format of the data presented in Chapter 1:
Each variable forms a column.
Each observation forms a row.
Each value must have its own cell.
When data are stored in this format, it is called tidy data(Wickham 2014).
Figure 7.2: Tidy data format. source: Wickham and Grolemund (2016)
The R code below imports two versions of the data (a tidy and non-tidy version). The R function read.csv() is described in Appendix C. The functionhead() (tail()) will print out the first (last) few rows of a data frame.
TipR functions
### head( data , n )
# data: Replace with the name of the data frame
# n: Set equal to a number. The first 'n' rows will be printed
#
### tail( data , n )
# Note: It will print the last 'n' rows.
#
Note
The code below imports the data (tidy and non-tidy format) for the case study given in Chapter 4.
# Import tidy data from a csv file and store it in a data frame called "bmidf"#bmidf <- read.csv( "datasets/BMIcsdata.csv" )library(CSUBstats)data(BMIcsdata)bmidf<-BMIcsdata#load the dataset named `BMIcsdata` into your current workspace# Import non-tidy data from a csv file and store it in a data frame called "bmidfnt"#bmidfnt <- read.csv( "datasets/BMIcsdataNT.csv" )library(CSUBstats)data(BMIcsdataNT)bmidfnt<-BMIcsdataNT#load the dataset named `BMIcsdataNT` into your current workspace# View the first 5 rows of the "bmidf" data frame.head(bmidf , n=5)#> Country Sex Region Year BMI#> 1 Afghanistan Men Rural 1985 19.7#> 2 Afghanistan Men Urban 1985 22.4#> 3 Afghanistan Men Rural 2017 22.5#> 4 Afghanistan Men Urban 2017 23.6#> 5 Afghanistan Women Rural 1985 20.1# View the last 5 rows of the "bmidfnt" data frame.tail(bmidfnt , n=5)#> Country Sex Region yr.1985 yr.2017#> 788 Zambia Women Urban 22.5 25.5#> 789 Zimbabwe Men Rural 20.9 21.9#> 790 Zimbabwe Men Urban 22.1 23.3#> 791 Zimbabwe Women Rural 24.0 24.5#> 792 Zimbabwe Women Urban 26.5 27.1
The tidy format makes a clear distinction between variables, observations, and values. The second data set may be useful for viewing year-to-year changes, but it places year information in both rows and columns. Tidy data make variables and their observed values explicit.
7.5 Codebooks
An important task in a data based research project is to document the structure and content of project dataset. This can be done in the form of a codebook, a file that lists at least the names of the variables, variable descriptions, and data source. Sometimes it may also provide the variables’ type (e.g., continuous, categorical) and levels of factor variables (e.g., 1 = disagree,…, 3 = agree). Writing a codebook is an important step in managing the data aspect of your research project. The codebook will serve as an important reference for the researcher and any collaborators, and there are multiple approaches for creating a codebook, with Excel being a popular choice. The next section contains a sample codebook for the case study data.
Note
Here, a codebook is provided for the data given in Chapter 1.
Codebook
The Delta smelt case study data frame has the following dimensions:
Feature
Result
Number of observations
20
Number of variables
3
Codebook summary table
Label
Variable
Class
# Unique values
Missing
Description
low dosage (4.5 micromols per square meter per second), medium dosage (6.75 micromols per square meter per second), high dosage (9 micromols per square meter per second)
[Light]
factor
3
0.00 %
low dosage (2 NTUs), medium dosage (5.5 NTUs), high dosage (9 NTUs)
[Turbidity]
factor
3
0.00 %
Percentage of early-stage Delta smelt larvae (0-40 days post hatch) that survived
[Survival]
numeric
20
0.00 %
Variable list
Light
low dosage (4.5 micromols per square meter per second), medium dosage (6.75 micromols per square meter per second), high dosage (9 micromols per square meter per second)
Feature
Result
Variable type
factor
Number of missing obs.
0 (0 %)
Number of unique values
3
Mode
“low”
Reference category
low
Observed factor levels: “high”, “low”, “med”.
Turbidity
low dosage (2 NTUs), medium dosage (5.5 NTUs), high dosage (9 NTUs)
Feature
Result
Variable type
factor
Number of missing obs.
0 (0 %)
Number of unique values
3
Mode
“low”
Reference category
low
Observed factor levels: “high”, “low”, “med”.
Survival
Percentage of early-stage Delta smelt larvae (0-40 days post hatch) that survived
Feature
Result
Variable type
numeric
Number of missing obs.
0 (0 %)
Number of unique values
20
Median
59.56
1st and 3rd quartiles
55.79; 67.24
Min. and max.
45.02; 89.09
Report generation information:
Created by: Eduardo L. Montoya (email:emontoya2@csub.edu).
Source: The data are a subset of the data provided by Dr. Tien-Chieh Hung. The data are a subset of the data analyzed in Tigan et al. (2020).
Codebook created using dataMaid v1.4.1 [Pkg: 2021-10-08 from CRAN (R 4.1.3)]
Tigan, Galen, William Mulvaney, Luke Ellison, Andrew Schultz, and Tien-Chieh Hung. 2020. “Effects of Light and Turbidity on Feeding, Growth, and Survival of Larval Delta Smelt (Hypomesus Transpacificus, Actinopterygii, Osmeridae).”Hydrobiologia 847 (13): 2883–94. https://doi.org/10.1007/s10750-020-04280-4.
Wickham, Hadley, and Garrett Grolemund. 2016. R for Data Science: Import, Tidy, Transform, Visualize, and Model Data. " O’Reilly Media, Inc.".
There are various functions for importing data from different file formats. While some of these functions may correctly treat factor variables as factors, the functions used in these modules for data importation do not do so by default.↩︎
