ITEM ANALYSIS
Computation and examination of any statistical property of an item response distribution
Item parameters fall into 3 categories/indices –
1. That describe distribution of responses to a single item
2. That describe degree of relationship between the response on the item and some criterion
of interest
3. That are a function of both, meaning relationship to item variance/mean and a criterion of
interest
Classical Test Theory (CTT)
Statistics used –
Correlation – measures how two variables are related
Covariance – how much two random variables vary together
Discrimination index – ability of test to discriminate between different levels of learning
Item difficulty – difficulty level of individual items
Reliability co-efficient – internal consistency of the test
Sample variance/SD – how spread out the scores are
Standard error of measurements – how much measured test scores are spread around a
true score
Item Analysis using CTT –
Item difficulty – For dichotomous items, its proportion of examinees who correctly answer
that item. For polylomous items, its evaluation of mean score. (if p=0.95 the test is very
easy, if p=0.35 item is difficult)
Item discrimination – Measured using point biserial correlation which correlates scores on
an item to scores on the total test. If the item is strong and measures the topic well, then
examinees who get the item right tend to score higher on the test (> 0.20). If its 0.00, item is
random data generator and worthless.
Difficulty Value (D.V.) Evaluations
D. V. Item Evaluation
0.20 – 0.30 Most Difficult
0.30 – 0.40 Difficult
0.40 – 0.60 Moderate Difficult
0.60 – 0.70 Easy
0.70 – 0.80 Most Easy
Discrimination Index (D.I.) Guidelines
D.I. Item Evaluation
>0.40 Very good items
0.30 – 0.39 Reasonably good but subject to improvement
0.20 – 0.29 Marginal items, need improvement
<0.19 Poor items, rejected or revised
Computation and examination of any statistical property of an item response distribution
Item parameters fall into 3 categories/indices –
1. That describe distribution of responses to a single item
2. That describe degree of relationship between the response on the item and some criterion
of interest
3. That are a function of both, meaning relationship to item variance/mean and a criterion of
interest
Classical Test Theory (CTT)
Statistics used –
Correlation – measures how two variables are related
Covariance – how much two random variables vary together
Discrimination index – ability of test to discriminate between different levels of learning
Item difficulty – difficulty level of individual items
Reliability co-efficient – internal consistency of the test
Sample variance/SD – how spread out the scores are
Standard error of measurements – how much measured test scores are spread around a
true score
Item Analysis using CTT –
Item difficulty – For dichotomous items, its proportion of examinees who correctly answer
that item. For polylomous items, its evaluation of mean score. (if p=0.95 the test is very
easy, if p=0.35 item is difficult)
Item discrimination – Measured using point biserial correlation which correlates scores on
an item to scores on the total test. If the item is strong and measures the topic well, then
examinees who get the item right tend to score higher on the test (> 0.20). If its 0.00, item is
random data generator and worthless.
Difficulty Value (D.V.) Evaluations
D. V. Item Evaluation
0.20 – 0.30 Most Difficult
0.30 – 0.40 Difficult
0.40 – 0.60 Moderate Difficult
0.60 – 0.70 Easy
0.70 – 0.80 Most Easy
Discrimination Index (D.I.) Guidelines
D.I. Item Evaluation
>0.40 Very good items
0.30 – 0.39 Reasonably good but subject to improvement
0.20 – 0.29 Marginal items, need improvement
<0.19 Poor items, rejected or revised
