Ordinal response model of community response

Abstract

Those analyzing community response to noise data often collapse ordinal scale data into a binomial variable to predict the percentage of the community that will be highly annoyed at given a noise dose. Collapsing the data simplifies the model, but also discards potentially useful information. An ordinal response model is one way to use all the information available in ordinal scale data. This paper compares the conventional binomial model to an ordinal response model to demonstrate the benefits and drawbacks of each. We have found that the ordinal response model may be a better option because some independent variables only influence response to noise at the lower end of the ordinal scale. The ordinal response model can detect effects across the entire ordinal range whereas the binomial model examines only the highly annoyed end of the scale.