Because you use only the discrimination index or discriminiation coeff. it is hard to make a selection on that base alone. I agree that it normally should be above 0.2 but 0.5 would be even better.
But the coeff and the index are both influenced by the P factor of a question. The reason for this is that a question that is very easy (p near 1(100) or hard p near 0 (0) means that allmost all students ( so the bad ones and the good ones) are scoring high (or low) on this question. In both cases there the question doesn't discriminate even if it is a great question (for that group it is just to easy or to hard).
We use all information we can get to decide if we want to skip a question. So I would advise the following.
1: Look at following parameters of the overall test
- average P ( is the test easy, average or hard)
- the reliability ( alpha, I am not sure if moodle does that). Is it above .7 than it is normally good enough.
- how many people took the test (more then 100 would be nice, more then 50 would be acceptable
- how many questions where there in the test (MC with 4 answers at least 40)
2: Then look at the individual questions.
- Look for high and low P values (P= 0,4---P=0,8 are the best questions) they are selecting
- Look for the disc. index/coeff (above .2 will be sufficient, but remember my first info)
- Look at the give answers from the questions who score not so well in P or disc.
- Kick out the questions who are really giving you any concern. You want it to be a selection instrument.
be carefull with the P value and disc. index when you have small groups. The effect will be stronger in small groups.
If you have only ten students and one good student give the wrong answer, your disc index and P value will drop big time.
As you see it is not as simple as you think. You should always look at the complete picture and be the expert who looks at the question.
Hope this epos will help you a bit further.