Novometric Analysis of Transition Matrices to Ascertain Markovian Order

Paul R. Yarnold

Optimal Data Analysis, LLC

The American National Election Panel Study modeled transitions in social class identification occurring between 1956, 1958, and 1960. Visual analysis suggested “…that transition probabilities linking class identifications in 1958 and 1960 vary with 1956 identification”. Transition tables were compared using Goodman’s chi-square procedure: chi-square=98.2, df=2, p<0.0001. Based on visual examination and non-disentangled omnibus chi-square findings, the null hypothesis of a first-order Markov model was rejected in favor of the alternative hypothesis that the underlying temporal process is a second-order Markovian (pp. 13-15). In contrast, novometric analysis indicated that a first-order Markov model is appropriate for these data.

View journal article