Здравствуйте. В конце статьи которую я читаю я не совсем понял последнюю часть. Сама статья описывает метод анализа категориальных данных (frequency tables), основываясь на стохастическом порядке, но последняя часть говорит о преимуществах такого анализа.
Не могли бы вы мне объяснить, что такое
.
We have shown in this paper that simple and intuitive ordinal considerations regarding stochastic ordering give rise to CA models of rank2. Within the rank2 model family ordinal models of uniform association of various degrees can be considered. Correspondence analysis models with rank higher than 2 which still use ordinal consideration (other than stochastic ordering) are thoroughly discussed and exemplified by Gilula & Haberman (1988). We would like to conclude this paper by briefly showing how rank 2 CA models can also be used for the analysis of variance of categorical data.
Light & Margolin (1971) and Gilula (1985) discuss issues of such analysis of variance in terms of measuring variability of categorical variables and the establishment of appropriate test statistics. Recall that if in the usual framework of the one-way analysis of variance,
denotes the
-th observation in the
-th sample, then the relevant sum of
squares are
where
is the overall mean,
is the mean in the
-th sample and
is the size of the
sample drawn from the jth population. In our context we have J-multinomial populations,
all defined on the same I-category random variable
. Here
is the size of
the sample drawn from the jth population, and nii is the frequency in the
-th cell. Now
assume that the rank 2 CA model holds.