Question regarding the DM model for polytomous items‏‏: reply from Dr. van der Ark

一個多禮拜前寫信問Dr. van der Ark  的問題，終於接到回信了，能接到作者(應該是大師)的回覆，真的蠻感動的!!

我的問題內容大致如下:

...   My question  is about how to deal with different numbers of response categories in a given polytomous data set. In my questionnaire, there are six items scored on a 3-level scale (0, 1, 2 points) and two items are scored on a 4-level scale (0, 1, 2, 3points). I would like to confirm if it is still all right to use Mokken scale analysis to analyze my data. Should I collapse the response categories, e.g. 4-level scale into 3-level scale to make the same format?

 There was inconclusive evidence on multiple studies for whether or not it is okay to use different numbers of response categories in a give set.   ....

 

Dr. van der Ark的回覆如下:

"The problem with different numbers of response categories is that that is unclear what the scale score means. Scores 0, 1, and 2 in the first set of items are not equivalent to the scores 0, 1, and 2 in the second set of items because in the latter set there is also a category 3.

So I would advise against having scales consisting of items with different numbers of categories, especially for investigating IIO where the interest lies in ordering of items by the mean item score. You may collapse the response categories if you believe that is meaningful; that is, if you believe that by collapsing the scores in the first set of items the scores in both sets of items are comparable."
 

就Dr. van der Ark的回覆，是建議使用collapsing the scores，但對於我們來說，

項目選項的meaning可能就會不見，因為目前的選項對於item來說有其意義 ; 但從另一個

角度來看，如果說軟體無法(或本意並無)可計算不同response categories，那目前出來

的結果該如何解釋呢?!! 思考中~~