CODEWITHSUNDEEP
rcorrelation
Haley
Haley

Reputation: 159

Why am I only getting 1 and -1 from the cor() function in R?

I'm trying to perform a Pearson correlation with the cor() function, but the output only gives me 1 and -1, not the coefficient itself. So when I go to plot the matrix with corrplot(), I only see those 1 and -1 values. How do I fix this? My dataset can be found here, and see my script down below: 

##Must load the libraries we will need! IF you have not installed the packages, do that before you start. 
library("corrplot")
##Load in your datasets 
D1=BPT5test
##if you don't have a Y (i.e, you want the same thing to be in both axis), leave this blank
D2=
##Run the spearman correlation. If you want to do a Pearson, change "spearman to "pearson"
##If you have 0s in your dataset, set use = "complete.obs", if you have no 0s, set use = "everything"
CorTest=cor(D1, use = "everything", method = "pearson")
##Let's get to plotting! 
##Lots of changing you can do! 
#Method can be "circle" "square" "pie" "color"
#ColorRampPalette can be changed, "blue" being the negative, "White" being '0', and "red" being the positive
#Change the title to whatever you want it to be 
#tl.col is the color of your labels, this can be set to anything.. default is red
CorGraph=corrplot(CorTest, method = "circle", col = colorRampPalette(c("blue","white","red"))(200), title = "Pearson's Correlation of High-Fat Sugar at 8 weeks", tl.cex = .5, tl.col = "Black",diag = TRUE, cl.ratio = 0.2)

Upvotes: 2

Views: 2037

Answers (2)

Vandenman
Vandenman

Reputation: 3176

Your dataset contains only 2 observations per variable. The correlation between any two variables consisting of only two observations is always -1 or 1. To see for yourself, try running replicate(1e2, cor(rnorm(2), rnorm(2))) which calculates 100 correlations between two variables consisting of two observations. The result is always -1 or 1.

Upvotes: 3

aldo_tapia
aldo_tapia

Reputation: 1228

It's because you have only two observations by column.

test <- data.frame(a=c(1,2),b=c(2,3),c=c(4,-2))
cor(test, use = "everything", method = "pearson")
   a  b  c
a  1  1 -1
b  1  1 -1
c -1 -1  1

You can't expect a different output with only two values, check Pearson correlation formula.

Since three or more you will have more variation:

test <- data.frame(a=c(1,2,3),b=c(2,3,5),c=c(4,-2,-10))
cor(test, use = "everything", method = "pearson")
           a          b          c
a  1.0000000  0.9819805 -0.9966159
b  0.9819805  1.0000000 -0.9941916
c -0.9966159 -0.9941916  1.0000000

Upvotes: 2

Related Questions