3

Box-Cox, log and arcsine transformations have the aim of make the data more Normal. My question is: how can I choose between each one of these transformations? Which assumptions do I need to have to do that?

Normally I see people trying these transformations when modeling and simply picking the one which better performs - without any strong assumption or theory behind that.

I'm not so versatile in statistics but I'm trying to further understand it.

kjetil b halvorsen
  • 77,844
Rods2292
  • 371
  • 1
    (1) The log is a Box-Cox transformation. (2) There's plenty of theory: see the original papers and textbooks on EDA c. 1970-1990. For an example, see our thread at https://stats.stackexchange.com/questions/10975 and threads on Box-Cox transformations at https://stats.stackexchange.com/search?tab=votes&q=Box-cox. – whuber Apr 10 '18 at 20:52
  • Do you have some context? This is to unspecific for an answer, see this long list of similar posts. As for the arcsine https://stats.stackexchange.com/questions/20772/are-ecologists-the-only-ones-who-didnt-know-that-the-arcsine-is-asinine – kjetil b halvorsen Oct 02 '20 at 16:55

0 Answers0