Angoff and Hofstee methods

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

group_theory

EX-TER-MIN-ATE!'
Staff member
Administrator
Volunteer Staff
Lifetime Donor
20+ Year Member
Joined
Oct 2, 2002
Messages
4,863
Reaction score
2,248
I know most of you guys/gals have never heard of the Angoff and the Hofstee methods but they are an integral part of your lives (exam wise). Just a little FYI

The Angoff method begins with a discussion of the candidate who is minimally qualified to pass the exam.
Panelists then review each test item and estimate the probability this minimally qualified candidate will
answer the item correctly. A passing score is derived for each panelist by summing these item probabilities.
The final passing score is calculated by taking the average passing score across panelists.

The Hofstee method does not require individual item judgments. Instead, panelists are asked to give their
impressions of what the minimum and maximum failure rates should be for the exam, as well as what the
minimum and maximum percent correct scores should be. These minimum and maximum failure rates and
percent correct scores are averaged across panelists and projected onto the actual score distribution to
derive a passing score. The Hofstee method is often used to evaluate or adjust the passing score derived
using the Angoff method. When the two methods produce similar passing scores, they validate each other.
When they diverge, executive committees typically use both sources of information to decide where to set
the passing score.

The Angoff method is the most popular method used to set passing scores on licensure and certification
exams.

Members don't see this ad.
 
One shudders to think of the unverifiable assumptions, outright guesswork, and fortune-telling that must go into trying to use either of these "methods."

How does one define "minimally qualified?" Is this based soley upon med school grades?
Does it assume no review is done by these "minimally qualified people?"
How is this "probability" measured?"

If this resembles the kind of "science" that goes into such things as economic forecasts, it can be safely ignored!
 
For the USMLE, they use two seperate groups for each step. There are around 12 panelists for each group. The standard setting panelists are selected from a variety of
disciplines and at least one D.O. sat on each panel. Efforts are made to achieve sex and racial/ethnic equity
within the standard setting panel. About 20-25% of the panelists for the Level 2 and 3 studies were non-Caucasian. The percentage of female standard setting panelists ranged from 17% to 45% for each Step. Medical specialty area and geographic region are also used as criteria for selecting standard setting panelists.

The panelists provide initial Angoff ratings on a representative subset of exam items (about 200 items) and are then given the item statistics as well as an opportunity to revise their ratings. After providing their final item ratings, the panelists also provide Beuk/Hofstee ratings. The final standards are selected using (a) the results from two independent Angoff panels, (b) an
analysis of examinee performance trends, and (c) an analysis of survey data gathered from various
constituencies.

Note: This info is from the FSMB, along with the use of copy+paste for the majority of this post along with the first post of this thread (my plagarism disclaimer)
 
Top