Item Analysis

 

Item analysis:     “refers to the computation and examination of any statistical property of examinees’ responses to an individual test item”.

 

Three types of indices

1.  Distributional indices (descriptive stats: mean, mode, median, variance, range, ...)

2.Discrimination indices (relationship between the response to an item with a criterion)   

                 2.1 Index of discrimination

                 2.2 Point-biserial correlation

                 2.3 Biserial correlation

                 2.4 Item-total correlation (SPSS)

                 2.5 Phi correlation

                 2.6 Tetrachoric correlation 

3. Adjusted discrimination

                 3.1 Item reliability


 

CATEGORY 1:  Distributional indices

 

Indices evaluated:         frequencies, mean, mode, median, variance, range, etc...

Basic principles

1.      Low variance or skewed data (resulting from restriction of range) => possible attenuation

2.      Normally distributed data => No attenuation

 

Dichotomous data

Item mean or item difficulty or item endorsement

 

Pj

σ2j = pjqj, where qj=1-pj

.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
1.00

.00
.09
.16
.21
.24
.25
.24
.21
16
.09
.00

 

Next