I have a dataset of variables that fail to conform to a normal distribution (Kolmogorov-Smirnov sig. value = 0 for all the variables).

In order to perform parametric statistics on these variables I need to make them have a normal distribution. I have already tried $log_{10}$ transformation and sort transformation but the distribution remains skewed. I read about the Box-Cox method to transform a variable but I'm not familiar with it. Can anyone briefly explain me the logic behind it and how to use it in SPSS?

  • $\begingroup$ You can get an explanation of the box-cox transformation here: en.wikipedia.org/wiki/Power_transform $\endgroup$ – kjetil b halvorsen Dec 5 '14 at 17:29
  • 2
    $\begingroup$ If you use your data to (1) test for normality and then (2) identify a transformation of the variables, and then (3) apply your normal-based statistical procedures on the transformed data, then (1) the p-values will be a little too low and (2) you likely won't be testing the null hypotheses you intended. (For instance, a t-test of means based on logarithms is actually a test of the geometric means of the original data.) Moreover, many Normal-theory parametric tests do not require the data look normally distributed. So maybe you don't have a problem at all. $\endgroup$ – whuber Dec 5 '14 at 17:59
  • $\begingroup$ But if you do need to do Box-Cox, it is available in the automatic data prep procedure (you can turn all the other parts off). $\endgroup$ – JKP Dec 6 '14 at 22:49

Your Answer

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

Browse other questions tagged or ask your own question.