Ordinal and Multinomial models

1. Ordinal and Multinomial models

2. Ordinal Outcomes 3 or more categorical outcomes, which can be treated as ordered Bond ratings (AAA, AA, � B, C, �) Likert scales (e.g. responses on a 1-7 scale, from strongly disagree to strongly agree) Often analyzed as continuous Can use logit or probit That was a not-so-quick review of binary models. Now we�re ready for ordinal models. [Explain what�s on the slide.] Statistical packages vary as to how these ordered outcomes are coded � typically using consecutive integers. For this talk, I�ll assume we�re analyzing a Likert scale starting with 1, so outcome 1 corresponds to the lowest ordered category, 2 the next, and so on. Analyzing as continuous assumes it�s interval data, I.e. that the distance from 1 to 2 is the same as the distance from 2 to 3, etc. But this may not be true if the categories are e.g. strongly disagree, disagree, neutral. That was a not-so-quick review of binary models. Now we�re ready for ordinal models. [Explain what�s on the slide.] Statistical packages vary as to how these ordered outcomes are coded � typically using consecutive integers. For this talk, I�ll assume we�re analyzing a Likert scale starting with 1, so outcome 1 corresponds to the lowest ordered category, 2 the next, and so on. Analyzing as continuous assumes it�s interval data, I.e. that the distance from 1 to 2 is the same as the distance from 2 to 3, etc. But this may not be true if the categories are e.g. strongly disagree, disagree, neutral.

3. Ordinal logit