2
$\begingroup$

Everything I've read about degrees of freedom says more is better - it is an indication of a model's flexibility.

I'm now running a power analysis in R and I've noticed that lower degrees of freedom give higher power.

Example 1 (DF = N - 1):

library(pwr)

pwr.chisq.test(w = 0.7, N = 30, df = 29, sig.level = 0.05)

     Chi squared power calculation 

              w = 0.7
              N = 30
             df = 29
      sig.level = 0.05
          power = 0.5115032

NOTE: N is the number of observations

Example 2 (DF = 1):

pwr.chisq.test(w = 0.7, N = 30, df = 1, sig.level = 0.05) 

     Chi squared power calculation 

              w = 0.7
              N = 30
             df = 1
      sig.level = 0.05
          power = 0.9695413

NOTE: N is the number of observations

I assume in this case DF is supposed to refer to the number of parameters? If that's correct, why is it also used as sample size minus parameters?

$\endgroup$
1
  • $\begingroup$ What exactly did you learn about degrees of freedom and where? It definitely is not the case that "always more is better", so the claim is not correct. $\endgroup$
    – Tim
    Commented Feb 16, 2022 at 14:13

1 Answer 1

0
$\begingroup$

What software are you using? Or more exact how did you compute the power.

You can check here: https://www.statskingdom.com/33test_power_regression.html

that increasing you sample size, increases you power.

$\endgroup$
1
  • $\begingroup$ I used pwr.chisq.test() from the package "pwr" in R. I'll add my calculation to the original question. $\endgroup$
    – Picapica
    Commented Feb 16, 2022 at 16:00

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.