I would like to perform a stepwise regression on a number of continuous independent variables to determine which ones best predict the dependent variable. The dependent variable is ordinal in nature with 3 categories. How can I do this, or do I have to somehow just dichotomize my DV so I can do a logistic stepwise regression?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
6
Regarding the model: Don't make dummies out of your ordinal dependent. You need to use an ordinal logistic regression model. Its hard to fully answer without more details on your data or which statistical package you use. If your dependent was categorical you would use a multinominal logistic regression model. This is a decent tutorial on fitting and interpreting the ordinal model in R. Edit: Ordinal logistic regression with SAS, and Interpreting ordinal logistic output in SAS.
Regarding stepwise regression: Note that in order to find which of the covariates best predicts the dependent variable (or the relative importance of the variables) you don't need to perform a stepwise regression. You need standardized coefficients. In R you can do it using the scale() function on your data set, but all statistical packages have equal (or easier) mechanisms. Comparing the size of the standardized coefficients will give you the answer. Using stepwise regression will help you understand which model is most economic in that it incorporates only those which benefit the model. However, it is not a very recommended method as it may not find the best model. You might prefer to use theoretical considerations.
Edit: regarding explained percent variance: If the previous method of finding relative importance is not good enough and you need the explained percent of the variance per variable, you are sadly out of luck. $R^2$ does not exist for logistic models. See This explanation for more details on pseudo $R^2$ From the UCLA stat help (from which all links here are taken):
The model estimates from a logistic regression are maximum likelihood estimates arrived at through an iterative process. They are not calculated to minimize variance, so the OLS approach to goodness-of-fit does not apply. However, to evaluate the goodness-of-fit of logistic models, several pseudo R-squareds have been developed. These are "pseudo" R-squareds because they look like R-squared in the sense that they are on a similar scale, ranging from 0 to 1 (though some pseudo R-squareds never achieve 0 or 1) with higher values indicating better model fit, but they cannot be interpreted as one would interpret an OLS R-squared and different pseudo R-squareds can arrive at very different values
answered Oct 18, 2016 at 17:02
-
$\begingroup$ Thank you! My data consists of some continuous IVs (cholesterol levels) and my DV is a self-reported symptom score (none - mild/mod - severe). I am interested in determining which factors independently are associated with the score, using % variance of each in contribution to the DV. Would an ordinal stepwise logistic be appropriate? Also, I am using SAS. $\endgroup$– LucyCommented Oct 18, 2016 at 17:43
-
-
$\begingroup$ Thank you so much for your response. I decided to combinate mild/mod and severe so we have a binary logistic regression instead of the ordinal. Your comments are still very applicable thank you again. Most boards seem to suggest against using stepwise. It seems the lesser of all the evils might be backward elimination, so I have decided to go with that. If it is ok may I continue to ask for your help? My dataset has 15 vs 60 observations for my 0:1 DV, respectively. Given the small 0 dataset, i feel i should not take out a validation sample, will this be a problem? $\endgroup$– LucyCommented Oct 19, 2016 at 7:34
-
$\begingroup$ Actually, going back over my notes it seems the rule of thumb is to have at least 10 observations of each outcome (0 and 1) per each possible predictor! My sample size is too small!! What can I do? $\endgroup$– LucyCommented Oct 19, 2016 at 7:50
-
1$\begingroup$ It is preferable to open a new question for this, as it will be better answered. Your sample size indeed seems small, as you might not have enough variance per category to encompass the different values of all of your predictor variables. How many do you have? $\endgroup$ Commented Oct 20, 2016 at 8:02