# Regression Model when dependent variable is ranking

I am trying to construct a regression model which the value would be the ranking of the products. There is 9 total in ranks, where 1 means the top 1.I am wondering which model should I use.

Thank you so much!

A logistic regression model is suited for this scenario.

This model is one which has an ordinal variable (your ranking scale for the products) as opposed to an interval variable where intervals between each value are equal in distance.

If you are using R, you can use the glm model in order to run this regression.

Assuming y = product ranking, X1 = Independent Variable 1, X2 = Independent Variable 2:

regression<-glm(y ~ X1 + X2, data=yourdata) is the command you would use in R to run this regression.

A few points to note with this model:

• According to Studenmund (2010), a logistic regression model will generally show more accuracy in terms of significance when the number of observations are 500 or above. This is because a lower degree of variation in the dependent variable may not capture the true significance of the correlation with a lower number of observations.

• The coefficients are interpreted as odds ratios - not in the same way as in a standard OLS regression model. As an example, this page by UCLA demonstrates that an odds ratio can be calculated by raising e to the power of the coefficient in your regression model.

Hope this helps.

• This way he/she will lose all information relative to ordering of the output rank. it's much better to use ordinal logistic regression: stats.idre.ucla.edu/r/dae/ordinal-logistic-regression May 14, 2017 at 22:15
• Correct, the ordering is important. The above example being more of a guide to how the logistic regression could be set up more generally. May 14, 2017 at 22:53
• Aside from the above comments about ordinal logistic regression being the right model, to make the above model logistic, don't you need to add family=binomial(link=logit) to the call to glm? Sep 9, 2017 at 19:39

You can use an ordinal logistic model.