[Rasch] What is the relationship between number of items and the number of participants needed?
johnbarnard at bigpond.com
Wed Jun 19 09:36:36 EST 2013
As much as I like rules of thumb, one needs to be careful of course. The
only categorical answer I can think of is "It depends." The higher the
stakes ... It is possible to calibrate with few people and items but the
lack of robustness will make it dangerous to interpret. Why do it if it is
useless? I would think around at least 100 people, but can drop the number
of items to say 20 in general. Now if we only had 65 people ... Likewise
with the rating scales. Often we get few people at the upper and lower ends
and then we may consider relaxing the 10 observations in these categories.
But, it depends... As Ben used to say - is it practically of use and does it
From: rasch-bounces at acer.edu.au [mailto:rasch-bounces at acer.edu.au] On Behalf
Of Mike Linacre
Sent: Wednesday, 19 June 2013 9:13 AM
To: rasch at acer.edu.au
Subject: Re: [Rasch] What is the relationship between number of items and
the number of participants needed?
You asked: "What is the relationship between number of items and the
number of participants needed?"
Here are a few guidelines to launch our discussion:
1) at least 30 observations of each item and of each person, if possible.
2) at least 10 observations of each score-level on each rating scale.
3) since we are usually more concerned about item fit than person fit,
there must be more persons than items.
Folks, please tell us about your own guidelines ....
Rasch mailing list
Rasch at acer.edu.au
More information about the Rasch