I have collected linguistic data from one participant over seven months (one text a month) to explore lexical and syntactic development in second language writing. Her written samples were coded for type-token ratio (TTR) and dependent clause per clauses ratio (DC/C).

TTR: (0.652, 0.579, 0.616, 0.521, 0.58, 0.758, 0.718) DC/C: (0.41463, 0.36, 0.3871, 0.5, 0.33333, 0.40909, 0.34483)

I would like to see whether the relationship between the two indices is negative or positive.

What statistical test should I use to see how the two variables correlate?

Many thanks in advance.


1 Answer 1


First, plot your data. Always plot your data.

> TTR <- c(0.652, 0.579, 0.616, 0.521, 0.58, 0.758, 0.718)
> DCC <- c(0.41463, 0.36, 0.3871, 0.5, 0.33333, 0.40909, 0.34483)
> opar <- par(mfrow=c(2,1))
> plot(TTR,type="o",pch=19)
> plot(DCC,type="o",pch=19)
> par(opar)


After that, you can run a correlation between your data and the observation index (or the number of days since the first measurement, if this is different for different intervals). The "normal" Pearson correlation assumes a linear relationship, while Spearman and Kendall look for "general up-and-down" relationships. This question may help in deciding which to use: Kendall Tau or Spearman's rho?

> cor.test(TTR,seq_along(TTR),method="pearson")

        Pearson's product-moment correlation

data:  TTR and seq_along(TTR)
t = 1.2264, df = 5, p-value = 0.2747
alternative hypothesis: true correlation is not equal to 0
95 percent confidence interval:
 -0.4267024  0.9058875
sample estimates:

> cor.test(TTR,seq_along(TTR),method="kendall")

        Kendall's rank correlation tau

data:  TTR and seq_along(TTR)
T = 13, p-value = 0.5619
alternative hypothesis: true tau is not equal to 0
sample estimates:

> cor.test(TTR,seq_along(TTR),method="spearman")

        Spearman's rank correlation rho

data:  TTR and seq_along(TTR)
S = 32, p-value = 0.3536
alternative hypothesis: true rho is not equal to 0
sample estimates:

In this particular case, none come out significant in a two-sided test (nor would they be significant in a one-sided test). To be honest, the plots don't really suggest a clear slope in one or the other direction.

  • $\begingroup$ Many thanks for your answer Stephan. I was interested to see how the two variables, namely the TTR and the DC/C indices, correlate with each other and not with time. Now I can perform that statisitcal analyses in R. $\endgroup$
    – A. Wind
    Commented Oct 27, 2017 at 12:04
  • $\begingroup$ Ah, sorry, then I misunderstood. But the concept would be similar. (And since you do have a time dimension to your data, I'd recommend that you look at temporal dynamics, too.) $\endgroup$ Commented Oct 27, 2017 at 14:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.