How to calculate correlation between multiple variables in R?

Understanding Correlation in Statistical Analysis

In statistical analysis, understanding the relationships between different variables is fundamental. A primary tool for quantifying the strength and direction of a relationship between two variables is the Pearson correlation coefficient (r). This metric specifically measures the degree of linear association between variables, assuming a continuous scale and a roughly normal distribution. Calculating correlations effectively is essential for modeling, feature selection, and general data exploration in environments like R.

Interpreting the Pearson Correlation Coefficient

The Pearson correlation coefficient is bounded, always returning a value between -1 and 1. The sign and magnitude of this coefficient provide immediate insight into the nature of the relationship, allowing analysts to quickly gauge whether variables move in tandem or opposition, and how strongly that movement is correlated.

• -1: Indicates a perfectly negative linear association. As one variable increases, the other decreases consistently.
• 0: Indicates no discernible linear association. The variables are independent in terms of linear movement.
• 1: Indicates a perfectly positive linear association. As one variable increases, the other increases consistently.

Preparing the Data Frame for Correlation Analysis in R

This detailed tutorial will guide you through calculating correlations across various subsets of data using the statistical programming language R. We will utilize the built-in cor() function, which is highly versatile for generating correlation coefficients. To begin, we must first establish a sample data frame that contains multiple variables, representing different measurements or observations.

We construct a simple data frame named df, consisting of twelve observations across four numerical variables (a, b, c, and d). Notice how variables ‘a’, ‘b’, and ‘c’ appear to increase together, while variable ‘d’ shows a generally decreasing trend relative to the others. This structure will allow us to demonstrate positive, negative, and varying levels of correlation.

#create data frame
df <- data.frame(a <- c(2, 3, 3, 5, 6, 9, 14, 15, 19, 21, 22, 23),
                 b <- c(23, 24, 24, 23, 17, 28, 38, 34, 35, 39, 41, 43),
                 c <- c(13, 14, 14, 14, 15, 17, 18, 19, 22, 20, 24, 26),
                 d <- c(6, 6, 7, 8, 8, 8, 7, 6, 5, 3, 3, 2))

Example 1: Calculating Correlation Between Two Specific Variables

The most straightforward application of the cor() function is calculating the correlation coefficient for only two specific variables within the data frame. To achieve this, we simply pass the two vectors of interest—in this case, column ‘a’ and column ‘b’—as arguments to the function. This method is useful when performing preliminary checks on suspected relationships.

The syntax below utilizes the dollar sign ($) operator in R to extract the specific columns from the df object. The resulting output, 0.9279869, indicates a very strong positive linear correlation between variable ‘a’ and variable ‘b’.

cor(df$a, df$b)

[1] 0.9279869

Example 2: Analyzing Correlations Across a Subset of Variables

If the analysis requires calculating correlations among a specific group of three or more variables, the cor() function can accept a subsetted data frame or matrix as its argument. By passing a selection of columns, R automatically generates a correlation matrix showing the pairwise relationship between every combination of the selected variables.

In this example, we select columns ‘a’, ‘b’, and ‘c’ using vector notation c('a', 'b', 'c') within the indexing brackets df[,]. The resulting output is a 3×3 matrix where the diagonal values are 1 (representing a variable correlated with itself), and the off-diagonal values represent the specific Pearson correlation coefficient for that pair.

cor(df[, c('a', 'b', 'c')])

          a         b         c
a 1.0000000 0.9279869 0.9604329
b 0.9279869 1.0000000 0.8942139
c 0.9604329 0.8942139 1.0000000

Interpreting the symmetrical correlation matrix allows for rapid comparison of relationship strengths:

• The correlation between variables a and b is 0.9279869, indicating a strong positive relationship.
• The correlation between variables a and c is 0.9604329, representing the strongest positive relationship observed in this subset.
• The correlation between variables b and c is 0.8942139, also showing a high degree of positive correlation.

Example 3: Generating the Full Correlation Matrix

If the goal is to examine the relationships among all numerical variables present in a data frame, we can simply pass the entire data object (df) to the cor() function without any specific indexing. R automatically assumes that the user wishes to calculate the correlation matrix encompassing all columns. This is often the quickest way to get a complete overview of the dataset’s collinearity.

In the resulting 4×4 matrix below, we observe the relationships established in Example 2, but also introduce the coefficients involving variable ‘d’. Note that the correlations between variables a, b, and c, and variable d are consistently negative (e.g., -0.7915488 for a vs. d), confirming our initial observation that ‘d’ moves inversely to the other variables.

cor(df)

           a          b          c          d
a  1.0000000  0.9279869  0.9604329 -0.7915488
b  0.9279869  1.0000000  0.8942139 -0.7917973
c  0.9604329  0.8942139  1.0000000 -0.8063549
d -0.7915488 -0.7917973 -0.8063549  1.0000000

Example 4: Calculating Correlations with Mixed Data Types

When working with real-world datasets, data frames often contain variables of different types, such as factors, characters, or dates, alongside the required numerical variables. The cor() function requires numerical input and will return an error if non-numeric columns are included in the calculation. Therefore, it is crucial to subset the data to include only appropriate columns for correlation analysis.

To robustly handle mixed data types, we can use a combination of R functions to dynamically select only the numerical variables. The expression lapply(df, is.numeric) checks every column for its type, and unlist() converts the resulting list of TRUE/FALSE values into a vector suitable for indexing. This ensures a clean calculation regardless of how many non-numeric columns exist in the original data.

cor(df[,unlist(lapply(df, is.numeric))])

           a          b          c          d
a  1.0000000  0.9279869  0.9604329 -0.7915488
b  0.9279869  1.0000000  0.8942139 -0.7917973
c  0.9604329  0.8942139  1.0000000 -0.8063549
d -0.7915488 -0.7917973 -0.8063549  1.0000000

Example 5: Visualizing Pairwise Relationships using the Pairs Plot

While correlation coefficients provide precise numerical measures, visualizing the relationships between variables offers crucial context, helping analysts identify non-linear associations or outliers that might skew the Pearson correlation coefficient. A powerful tool for this purpose is the pairs plot, which displays scatter plots for every variable pairing, histograms for individual variable distributions, and often the correlation coefficient itself.

To generate a comprehensive pairs plot, we utilize the pairs.panels() function, typically found within the psych package in R. We must load this package first, and then pass our entire data frame (df) to the function. This visualization immediately confirms the strong positive correlations among a, b, and c, and the negative relationship involving d, as shown by the downward slope in those respective scatter plots.

#load psych package
library(psych)

#create pairs plot
pairs.panels(df)

Conclusion: Advanced Correlation Techniques in R

Mastering the calculation and interpretation of correlation coefficients is a core skill in data analysis using R. Whether calculating a single pairwise coefficient or generating a full correlation matrix, the cor() function provides a fast and reliable method for quantifying linear association. Furthermore, combining numerical correlation outputs with robust visualizations like the pairs plot ensures a thorough understanding of the underlying data structure and prepares the user for more complex multivariate statistical modeling.