Calculating and Visualizing Correlation Matrices for Multiple Variables in R

Basic Methods for Computing Correlation Matrices

In R, calculating the correlation between two variables typically involves the cor() function, with the basic syntax cor(var1, var2, method = "method"). However, when analyzing multiple variables (e.g., four or more), pairwise calculations become tedious and inefficient. Fortunately, R offers a more streamlined solution: applying the cor() function directly to a data frame containing multiple variables.

For example, using the built-in dataset VADeaths, a complete correlation matrix can be generated with the following code:

> cor(VADeaths)
             Rural Male Rural Female Urban Male Urban Female
Rural Male    1.0000000    0.9979869  0.9841907    0.9934646
Rural Female  0.9979869    1.0000000  0.9739053    0.9867310
Urban Male    0.9841907    0.9739053  1.0000000    0.9918262
Urban Female  0.9934646    0.9867310  0.9918262    1.0000000

The output is a symmetric matrix where diagonal elements are 1 (indicating perfect correlation of a variable with itself), and off-diagonal elements show correlation coefficients between different variables. This approach not only simplifies the code but also ensures consistency and accuracy in computations.

Handling Datasets with Discrete Variables

In practical data analysis, datasets often mix numeric and discrete (factor) variables. Directly applying cor() to such data frames may cause errors, as correlation calculations typically require numeric variables. To address this, we can first filter for numeric columns.

Using the mtcars dataset as an example—though it primarily contains numeric variables, demonstrating filtering remains instructive—numeric columns can be identified with lapply() and is.numeric():

> cor(mtcars[,unlist(lapply(mtcars, is.numeric))])
            mpg        cyl       disp         hp        drat         wt        qsec         vs          am       gear        carb
mpg   1.0000000 -0.8521620 -0.8475514 -0.7761684  0.68117191 -0.8676594  0.41868403  0.6640389  0.59983243  0.4802848 -0.55092507
cyl  -0.8521620  1.0000000  0.9020329  0.8324475 -0.69993811  0.7824958 -0.59124207 -0.8108118 -0.52260705 -0.4926866  0.52698829
disp -0.8475514  0.9020329  1.0000000  0.7909486 -0.71021393  0.8879799 -0.43369788 -0.7104159 -0.59122704 -0.5555692  0.39497686
hp   -0.7761684  0.8324475  0.7909486  1.0000000 -0.44875912  0.6587479 -0.70822339 -0.7230967 -0.24320426 -0.1257043  0.74981247
drat  0.6811719 -0.6999381 -0.7102139 -0.4487591  1.00000000 -0.7124406  0.09120476  0.4402785  0.71271113  0.6996101 -0.09078980
wt   -0.8676594  0.7824958  0.8879799  0.6587479 -0.71244065  1.0000000 -0.17471588 -0.5549157 -0.69249526 -0.5832870  0.42760594
qsec  0.4186840 -0.5912421 -0.4336979 -0.7082234  0.09120476 -0.1747159  1.00000000  0.7445354 -0.22986086 -0.2126822 -0.65624923
vs    0.6640389 -0.8108118 -0.7104159 -0.7230967  0.44027846 -0.5549157  0.74453544  1.0000000  0.16834512  0.2060233 -0.56960714
am    0.5998324 -0.5226070 -0.5912270 -0.2432043  0.71271113 -0.6924953 -0.22986086  0.1683451  1.00000000  0.7940588  0.05753435
gear  0.4802848 -0.4926866 -0.5555692 -0.1257043  0.69961013 -0.5832870 -0.21268223  0.2060233  0.79405876  1.0000000  0.27407284
carb -0.5509251  0.5269883  0.3949769  0.7498125 -0.09078980  0.4276059 -0.65624923 -0.5696071  0.05753435  0.2740728  1.00000000

This method ensures that only numeric variables are included in the correlation calculation, preventing errors due to data type mismatches. In real-world applications, this step is crucial if the dataset contains genuine factor variables.

Visualization and Advanced Analysis of Correlations

Beyond numerical output, visualization tools can provide more intuitive insights into variable relationships. The pairs.panels() function from the psych package offers a comprehensive view:

library(psych)
pairs.panels(iris[1:4])  # Select the first four numeric columns

This function generates a scatterplot matrix with density plots on the diagonal and scatterplots with correlation coefficients off-diagonal, aiding in quick identification of linear relationships and distribution characteristics.

Another useful tool is the chart.Correlation() function from the PerformanceAnalytics package:

library(PerformanceAnalytics)
chart.Correlation(iris[1:4])

This chart not only displays correlations but also includes significance asterisks by default, enhancing the visualization of statistical inferences.

For more professional visualizations, the corrplot package provides rich options:

library(corrplot)
x <- cor(iris[1:4])
corrplot(x, type="upper", order="hclust")

By setting type="upper" and order="hclust", an upper triangular matrix is generated with hierarchical clustering, grouping highly correlated variables together for easier pattern recognition.

Statistical Testing and Extended Functionality

In some studies, correlation coefficients alone may be insufficient, and significance testing is required. The corr.test() function from the psych package provides this capability:

> corr.test(mtcars[1:4])
Call:corr.test(x = mtcars[1:4])
Correlation matrix 
       mpg   cyl  disp    hp
mpg   1.00 -0.85 -0.85 -0.78
cyl  -0.85  1.00  0.90  0.83
disp -0.85  0.90  1.00  0.79
hp   -0.78  0.83  0.79  1.00
Sample Size 
     mpg cyl disp hp
mpg   32  32   32 32
cyl   32  32   32 32
disp  32  32   32 32
hp    32  32   32 32
Probability value 
     mpg cyl disp hp
mpg    0   0    0  0
cyl    0   0    0  0
disp   0   0    0  0
hp     0   0    0  0

The output includes the correlation matrix, sample sizes, and p-values, offering complete information for hypothesis testing. For instance, all p-values are 0, indicating that these correlations are statistically significant at conventional levels.

In summary, R provides a comprehensive toolkit for analyzing multi-variable correlations, from basic computations to advanced visualizations and statistical tests. The core cor() function is simple and efficient, while extension packages enhance analytical and presentation capabilities. In practice, it is advisable to select tools based on research objectives: use cor() for quick insights, and combine visualization and testing packages for deeper analysis. Regardless of the method, these tools help researchers better understand relationship patterns in data, laying a solid foundation for subsequent modeling and decision-making.