0
$\begingroup$

I have 3 categorical independent variable and all the variables are more than 2 categories like 6 locations, 4 types, 6 maturity levels. Can I still use multiple regression? If so, how can I do that? If not, is there a better test for my experiment?

I was asked to clarify more: To explain more: I have 6 dependent variables and I want to see, for example if the location has an effect on my dependent variables. should I look at this effect one-by-one? Can I look the effect of 3 categorical independent variable on a dependent variable at the same time?

$\endgroup$
  • $\begingroup$ Re the edit: did you perhaps write "dependent" for "independent"? $\endgroup$ – whuber May 13 '18 at 20:06
  • $\begingroup$ I have 3 categorical independent variable and 6 independent variable as I wrote in the original question. $\endgroup$ – Lily May 18 '18 at 14:01
  • $\begingroup$ Again your comment is confusing. You wrote "6 dependent" in the question but now your comment says "6 independent"! $\endgroup$ – whuber May 18 '18 at 15:02
  • $\begingroup$ Sorry @whuber that was a typo! As stated in the original question, I have 3 categorical independent variable and 6 dependent variables and I am trying to explain the effect of these 3 categorical variables on each of those 6 independent variables. $\endgroup$ – Lily May 23 '18 at 21:40
  • $\begingroup$ I give up...once again you refer to those 6 variables both as "dependent" and "independent." $\endgroup$ – whuber May 24 '18 at 0:16
1
$\begingroup$

You can definitely use dummy variables (be sure to leave out one for your reference category), but the "maturity levels" sounds ordinal. Here's a related question with some interesting discussion about how to handle an ordinal predictor.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I am completely new so I am trying to learn. So let's say I am using dummy variables. Should the number of observations in the categories be same? I have read it somewhere but I might get confused since I am trying to learn it in a short time period. $\endgroup$ – Lily May 13 '18 at 12:52
  • 1
    $\begingroup$ There is no requirement for categories to be equally numerous. Nevertheless you're at the mercy of those small frequencies. The coefficient estimate inevitably is more sensitive to individual values. This can help or hinder. $\endgroup$ – Nick Cox May 14 '18 at 11:24
1
$\begingroup$

You can include dummy variables for each categorical variable or you can refer to multi-factor ANOVA tests (see e.g., this link)

| cite | improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ The same method in different guises! $\endgroup$ – Nick Cox May 14 '18 at 11:19

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