1
$\begingroup$

I have eight factors, four with 2 levels and four with 3 levels. These levels are not "high" and "low" in the sense that they can be turned off and on—but are instead things like "voiced" and "unvoiced", "male" and "female", or "front", "middle", and "closed."

Is it appropriate to arbitrarily choose which factor represents high and which represents low? Can I say when x = voiced = high = 1 and when x = unvoiced = low = -1, go through runs determined by a factorial design, and still get results that make sense? Are there any things I have to look out for when doing this?

My instinct is "yes, this is perfectly fine," but this type of analysis is fairly new to me, and I'd like to be certain.

$\endgroup$
1
$\begingroup$

If by "factorial design" you mean ANOVA then yes, you can do ANOVA with variables that do not have an order. You don't need the "high" and "low" and, indeed, shouldn't have them. E.g you can have a variable "voiced" and code it "unvoiced" and "voiced".

$\endgroup$
  • $\begingroup$ I suppose this is what I mean, yes. Thanks for your answer. Do you, by chance, have any suggestions for relatively quick reading material that would get me somewhat up to speed on ANOVA? I unfortunately don't have the time to read through a whole textbook, but am well versed in mathematics, so somewhat dense literature is ok. $\endgroup$ – AmagicalFishy Apr 5 '18 at 13:59
  • $\begingroup$ I don't have any particular links. There's a bunch of things to beware of in ANOVA (and in regression, which is the same model). $\endgroup$ – Peter Flom - Reinstate Monica Apr 5 '18 at 14:01
  • $\begingroup$ Ah, bummer. Thank you, though, for the answer! That is one step forward. I'll accept it when the 5-min. limit is up. $\endgroup$ – AmagicalFishy Apr 5 '18 at 14:02

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.