Ordinal Logistic Regression

Overview | How to | Example

 

 

Overview

Logistic regression is used to study the relationship between a response variable and its predictor(s), where the response variable is categorical. Ordinal Logistic Regression is used to perform logistic regression on a response variable that has 3 or more levels and these levels have a natural ordering. Examples of an ordinal variable is such as Agree/Neutral/Disagree and Small/Medium/Large/Extra Large.

Response variables and terms must be presented in columns. Covariates must be numeric and factors must have at least 2 distinct levels. The Model can include factors, covariates, interactions and nested term. Multiple terms must be separated by commas. Use * to represent interactions and () to represent nesting. The model must be hierachical, if an interaction term is included, all lower order interactions and main terms must also be included.

ProcessMA uses maximum likelihood estimates to model parameters.

 

 

How to

At the Excel Menu (For Excel 2007, go to Add-ins first)

  1. Choose ProcessMA > Statistics > Regression > Ordinal Logistic Regression

  2. In Response, select the column containing the response data

  3. In Term 1 to Term 10 , select the column containing data for the respective model terms and choose if the term is a covariate or a factor (Covariate: Numeric)

  4. In Model, enter the terms to be included in the model.

  5. In Link Function, choose the function to fit the response model

  6. Click OK

 

 

Example

You want to understand the effect of amount of pesticides and humidity on the life span of mosquitoes. You have classified life span as 1 (< 2 day), 2 (2 to 4 days) and 3 (4 to 6 days).

  1. Open data worksheet by choosing ProcessMA > Tools > Data

  2. Choose ProcessMA > Statistics > Regression > Ordinal Logistic Regression

  3. In Response, select AK - Lifespan

  4. In Term 1, select AL - Pesticide and choose Covariate

  5. In Term 2, select AM - Humidity and choose Factor

  6. In Model, enter Pesticide, Humidity

  7. Click OK

 

Results & Interpretation

From the logistic regression table, you can obtain the coefficients for the regression model. Constant(1) and Constant(2) are estimated intercept for lifepan 1 and 2 respectively. The p value for Humidity is 0.795 which indicates that there is insufficient evidence to conclude that humidity have an effect on mosquito population.

 

 

Copyright © ProcessMA