Posted by John Kleeman
I’m running a session at the Questionmark user conference next month on Item Analysis for Beginners and thought I’d share the answer to an interesting question in this blog.
When you run an Item Analysis report, one of the useful statistics you get on a question is its “p-value” or “item difficulty”. This is a number from 0 to 1, with the higher the value the easier the question. An easy question might have a p-value of 0.9 to 1.0, meaning 90% to 100% of participants answer the question correctly. A difficult question might have a p-value of 0.0 to 0.25 meaning less than 25% of participants answer the question correctly. For example, the report fragment to the right shows a question with p-value 0.998 which means it is very easy and almost everyone gets it right.
Whether such questions are appropriate depends on the purpose of the assessment. Most participants will get difficult questions wrong and easy questions right. In general, very easy and very difficult questions will not be as helpful as other questions in helping you discriminate between participants and so use the assessment for measurement purposes.
Here are three reasons why you might decide to include very difficult questions in an assessment:
- Sometimes your test blueprint requires questions on a topic and the only ones you have available are difficult ones – if so, you need to use them until you can write more.
- If a job has high performance needs and you need to filter out a few participants from many, then very difficult questions can be useful. This might apply for example if you are selecting potential astronauts or special forces team members.
- If you need to assess a wide range of ability within a single assessment, then you may need some very difficult questions to be able to assess abilities within the top performing participants.
And here are five reasons why you might decide to include very easy questions in an assessment:
- Answering questions gives retrieval practice and helps participants remember things in future – so including easy questions still helps reduce people’s forgetting.
- In compliance or health and safety, you may choose to include basic questions that almost everyone gets right. This is because if someone gets it wrong, you want to know and be able to intervene.
- More broadly, sometimes a test blueprint requires you to cover some topics that almost everyone knows, and it’s not practical to write difficult questions about.
- Easy questions at the start of an assessment can build confidence and reduce test anxiety. See my blog post Ten tips on reducing test anxiety for online test-takers for other ways to deal with test anxiety.
- If the purpose of your assessment is to measure someone’s ability to process information quickly and accurately at speed, then including many low difficulty questions that need to be answered in a short time might be appropriate.
If you want to learn more about Item Analysis, search this blog for other articles. You might also find the Questionmark user conference useful, since as well as my session on Item Analysis, there are also many other useful sessions including setting the cut-score in a fair, defensible way and identifying knowledge gaps. The conference also gives opportunity to learn and network with other assessment practitioners – I look forward to seeing some of you there.