2

I am trying to run a two-sample t-test for a difference between a treatment and control group. Data is not paired. When I subset my original dataframe, I found that I have unequal sample sizes (not an issue by hand, but R seems to make it an issue). Here is my code:

CG<-subset(data,treat=="Control")
TG<-subset(data,treat!="Control")
agep <-t.test(CG$age~TG$age)$p.value

The error I get is the following:

Error in model.frame.default(formula = CG$age ~ TG$age) : 
variable lengths differ (found for 'TG$age')

Yes! The lengths do differ. Not sure why that's a problem if I'm not running a paired test? Thanks in advance for any help.

Trying2Learn
  • 23
  • 1
  • 1
  • 3
  • 2
    You're using the formula interface incorrectly. Try a comma instead of ~ – Dason Oct 15 '18 at 07:34
  • Thank you! That was it! I can't believe I've been staring at this for hours over a comma. I appreciate it! – Trying2Learn Oct 15 '18 at 07:37
  • 1
    If you only have the two groups you could use the formula interface but you wouldn't want to subset your data first. t.test(age~treat, data=data) should do the same thing. – Dason Oct 15 '18 at 08:38

1 Answers1

11

If the unequal sample sizes are independent groups, then the mean can be parsed in R via an unpaired two-sample t-test.

First, ensure that your data pass a test of homoscedasticity--are the variances homogenous? We do this in R with a Fisher's F-test, var.test(x, y).

CG <- subset(data, treat == "Control")
TG <- subset(data, treat != "Control")
var.test(CG, TG)

If your p > 0.05, then you can assume that the variances of both samples are homogenous. In this case, we run a classic Student's two-sample t-test by setting the parameter var.equal = TRUE.

agep <- t.test(CG$age, TG$age, var.equal = TRUE)

If the F-test returns a p < 0.05, then you can assume that the variances of the two groups are different (heteroscedasticity). In this case, you can run a Welch t-statistic. Simply set var.equal = FALSE.

agep <- t.test(CG$age, TG$age, var.equal = FALSE)
Cory Caaz
  • 172
  • 2
  • 12