2
$\begingroup$

I'm newbie and bit lost... Can someone help me read this result from SPSS 1-sample Kolmogorov-Smirnov test (poisson distribution).

                                        SUBKEY1   SUBKEY2  SUBKEY3  SUBKEY4    SUBKEY5

N                                128       128       128      128    128

Poisson Parametera,,b     Mean          .4609     .4609    .4922     .5156  .4922

Most Extreme Differences  Absolute  .092      .092     .103      .113   .103

                      Positive  .079      .079     .088      .095   .088

                      Negative     -.092     -.092    -.103     -.113      -.103

Kolmogorov-Smirnov Z                1.037    1.037    1.171     1.276   1.171

Asymp. Sig. (2-tailed)              .233      .233     .129      .077   .129

From the result above, I define (in simple words):

Ho = satisfied
Ha = not satisfied

but I don't know how to specify the level significance for this test and I don't know whether have to read Asymp. Sig. (2-tailed) (p-value) or Kolmogorov-Smirnov Z (D value) to be compare with critical value...

$\endgroup$
4
$\begingroup$

If $\alpha = .05$ is conventional in your field, I would simply state, "A one-sample Kolmogorov-Smirnov test failed to reject the null hypothesis that the data followed the Poisson distribution (D = .092, .092, .103, .113, and .103 respectively for variables 1-5, N = 128 each, and p > .05 each)." You can also report that you set $\alpha = .05$ in your methods section. Note that $Z \approx \sqrt{n} D$ in the SPSS output and that the "Asymp. Sig." (p-values) in the output should be based on critical values of the Kolmogorov distribution to which $Z$ converges in distribution (see here for a description). However, you don't need to find or report these critical values -- I would just report the test statistics and p-value in the manner above.

$\endgroup$
  • $\begingroup$ Hi, lockedoff... tq for the answer... $\endgroup$ – blossom emerald Sep 9 '11 at 6:28
  • $\begingroup$ thanks for reply...i really appreciate that. But, i just want to make sure with what i understand that you state above...is it you compare the Dmax value (D = .092, .092, .103, .113, and .103) with the critical value with α=.05 (n>35 = 1.36/sqrt(n))? which is, if the D < critical value(0.120) than you fail to reject the null hypothesis? or you just compare the p-value>0.05? I'm still newbie in this field so, there a lot of question to be ask... =) $\endgroup$ – blossom emerald Sep 9 '11 at 6:46
  • $\begingroup$ I just want to make sure with what i understand that you state above...is it you compare the Dmax value (D = .092, .092, .103, .113, and .103) with the critical value with α=.05 (n>35 = 1.36/sqrt(n))? which is, if the D < critical value(0.120) than you fail to reject the null hypothesis? or you just compare the p-value>0.05? I'm still newbie in this field so, there a lot of question to be ask... =) $\endgroup$ – blossom emerald Sep 11 '11 at 15:04
  • $\begingroup$ Yes, based on a quick web search, it appears that the critical value for $D$ can be approximated by $1.36/\sqrt{n}$ for $\alpha = .05$. I would rather rely on the p-value reported by the program than an imprecise approximation, but either works. $\endgroup$ – lockedoff Sep 13 '11 at 17:16
  • $\begingroup$ It is correct if I've got higher value of p-value from 1-sample K-S its means i got better result than the other sample that i compared? eg: Variable_1 (p-value)= 0.09 and Variable_2 (p-value) = 0.06 means that variable_1 is better distribution than variable_2 or its just the same because that accept value is >0.05 so, there is no which result better than other? $\endgroup$ – blossom emerald Oct 18 '11 at 8:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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