How do you know whether to calculate the discrimination...

Name: Problem 2, Chapter 12: Measurement Theory in Action: Case Studies and Exercises
Uploaded: 2021-01-30T08:16:17-08:00
Duration: 1 min 24 s
Channel: Jeremiah Mbaria
Description: How do you know whether to calculate the discrimination index (which contrasts extreme groups), the biserial correlation, or the point-biserial correlation coefficient as your item discrimination statistic?

How do you know whether to calculate the discrimination index (which contrasts extreme groups), the biserial correlation, or the point-biserial correlation coefficient as your item discrimination statistic?

Key Concepts

Biserial Correlation

The biserial correlation coefficient is similar to the point-biserial but assumes that the dichotomous item responses arise from an underlying continuous distribution. It is particularly useful when the test item is thought to represent a cut-off on a continuous scale, thus providing an estimate that may reflect the true relationship with the underlying ability more accurately under certain psychometric assumptions.

Point-Biserial Correlation

The point-biserial correlation coefficient is a parametric statistic that measures the relationship between a dichotomous variable, such as a right or wrong response on an item, and a continuous variable, such as the total test score. It provides insight into how well the item score corresponds with the overall ability, and it is commonly used in item analysis when the responses are binary.

Item Discrimination

Item discrimination refers to the ability of an item on a test to differentiate between individuals who have a high overall ability and those with a lower ability. It is a key concept in psychometric analysis and is critical for determining which items contribute positively to the measurement of the underlying construct. High discrimination indicates that the item effectively distinguishes between different levels of ability.

Extreme Group Analysis

Extreme group analysis involves comparing the performance of test-takers at the upper and lower ends of the ability distribution. The discrimination index is a statistic derived from this method by contrasting the proportion of correct responses between these groups. It is useful when you want a non-parametric and simple measure of item performance, particularly in early stages of test analysis or when sample sizes are modest.

Transcript

Please give Ace some feedback