z***9
发帖数: 1052 | 1 有一个data， 是s x r的categorial data。 每一个combination有5种可能的
responses。现在要检验data是否是significant的，也就是data和每种response都是20
％的model是不同的。
我一下子能想到的方法是对每一种response都用chi square分开检验。但有没有方法可
以用一个model来检验5种response不是uniform distribution。 |
x******2
发帖数: 127 | 2 What do u want to test ?
I don't quite understand this sentence:
现在要检验data是否是significant的也就是data和每种response都是20
％的model是不同的。could you explain it a little bit more?
If u want to test if one predictor (predictors that have r levels or s
levels ) is significantly associated with responses , you might use MPH
model. |
x***x
发帖数: 3401 | 3 You can use multinomial logistic regression.
It can test if S and R is significant in predicting responses - if that is what
you are trying to get. |
