Hi,
WebPA normalises the group scores regardless of their size, so as far as the algorithm's maths is concerned there's no unfairness. At Loughborough, we usually recommend group sizes of 4 - 7. There's a general feeling that peer moderation with groups of 3 (or fewer!) members isn't worth it, and with larger groups the process is quite unwieldy and it's probably hard for any stand-out students to actually be recognised. Plenty of academics do use groups of 10+ though, it all depends how big your cohort is.
If a student drops out of a group completely, and the rest of the team score him/her down or give them zero overall, then everyone else in the team will benefit by having their grades inflated upwards. You could argue that the team deserves higher scores because they had to do proportionally more work being one man down, but the grade inflation can be quite dramatic and typically we take action to eliminate the effect.
The easiest way is to do this is to either:
(1) Identify the missing student before the assessment starts and remove them from the group.
(2) If the assessment is already underway, fix the group and create a separate assessment just for the group affected, so they can score just their contributing team mates.
(3) In rare cases, where the assessment timing means it's too awkward to reschedule a second run (e.g. Near Easter, Xmas, etc), we will manually remove the missing student and any associated scores from the assessment while it's running by editing the database directly.
--
Paul Newman
------------------------------------------------------
Senior PHP Developer
Centre for Engineering and Design Education
Loughborough University
------------------------------------------------------
From: WebPA [mailto:[log in to unmask]] On Behalf Of Clare Hopkins
Subject: Using Web-PA for groups of different sizes
I wonder if anyone could help me with a query about group sizes? Is the Web-PA algorithm affected when confronted by groups of different sizes (for example, group size ranging from 5 - 7 members or larger groups ranging from 10 - 12)? If this is the case, do you have any suggestions about how any unfairness can be rectified?
Also, what happens when one member who is listed as a member of the group, does not complete it because they have left the course? Does the algorithm take this into account?
|