Mann Whitney U test in R-CodePudding

I want to perform a Mann Whitney U test on a small data set of non-parametric data in R, can anyone help me find the right code? I have been trying to use the following code, to do the test straight from my data set, but I keep getting the error message: Error: Column.Heading.1 not found.

wilcox.test(Column.Heading.1, Column.Heading.2, data=DATA)

I'm new to stats so not sure if I'm missing something here. Thanks in advance!

CodePudding user response：

You might be assuming that the wilcox.test() will be able to find Column.Heading.1 and Column.Heading.2 inside DATA.

Unfortunately, this does not happen. The argument data only plays a role if you are giving a formula to the first element, i.e. Column.Heading.1~Column.Heading.2

If you want to use the x,y configuration that you are using you have to write the column in full, as if you trying to see the values on the console. For example.

wilcox.test(DATA$Column.Heading.1, DATA$Column.Heading.2)

Note that the formula and x,y configuration have different meanings. the formula assumes that Column.Heading.2 is a factor grouping the numbers (like "group1" or "group2"). the x, y configuration expects that both columns are filled with numbers.

CodePudding user response：

If you want to specify the data argument to wilcox.test then you need to use the formula interface, which means you will need a dataset in long format with a single column for the response and other for the group.

Otherwise to compare two columns of the same data frame you will need to refer to their locations explicitly. That is if we make a data frame:

dat <- data.frame(x=c(1.1,2.2,3.3),y=c(0.9,2.1,2.8))

Then we can run a Wilcoxon test either using:

wilcox.test(dat$x, dat$y)

with(dat , wilcox.test(x,y))