Skip to main content

An IERI – International Educational Research Institute Journal

Table 2 Bias and RMSE of Group Standard Deviation Estimates for DIF and nonDIF groups across the multigroup IRT models

From: Impact of differential item functioning on group score reporting in the context of large-scale assessments

 

Bias

 

RMSE

 

Unadjusted

Adjusted

Unadjusted

Adjusted

Direction

Percent

Size

Type

NonDIF

DIF

NonDIF

DIF

NonDIF

DIF

NonDIF

DIF

Positive

10%

Small

Non

–0.91

–0.67

–1.48

–1.03

5.04

4.97

5.06

4.73

   

Uni

–1.22

–1.53

–0.60

–1.16

4.97

5.42

4.95

5.29

  

Large

Non

–0.79

–0.03

–1.11

–0.47

5.38

5.13

4.92

4.99

   

Uni

–0.47

–2.17

–0.36

–0.84

5.16

6.95

4.75

5.28

 

20%

Small

Non

–1.04

0.32

–0.78

0.84

5.08

4.68

5.35

5.60

   

Uni

–1.38

–2.76

–0.09

–1.38

4.85

6.48

4.95

6.38

  

Large

Non

–1.12

0.62

–0.76

1.47

5.12

5.11

4.99

5.35

   

Uni

–0.98

–4.65

–0.60

–1.81

4.91

7.91

5.13

5.82

 

40%

Small

Non

–1.48

1.13

–0.79

1.30

5.07

5.11

4.82

5.35

   

Uni

0.20

–1.73

–1.04

–2.81

4.73

6.40

5.06

7.44

  

Large

Non

–1.22

2.20

–0.74

2.80

5.36

6.25

5.40

5.74

   

Uni

0.17

–5.83

–0.79

–8.19

5.31

10.16

5.28

13.95

Negative

10%

Small

Non

–0.42

–1.02

–1.07

–2.22

4.83

5.34

5.50

5.45

   

Uni

–0.76

–0.98

0.07

–0.49

4.79

5.63

5.22

5.22

  

Large

Non

–0.08

–3.03

–1.18

–2.50

4.94

6.03

4.72

5.38

   

Uni

–0.22

–1.86

–1.08

–1.54

5.02

6.27

5.38

5.64

 

20%

Small

Non

–0.61

–2.35

–0.12

–1.45

5.13

5.34

5.11

5.36

   

Uni

–0.71

–2.24

–1.33

–2.52

5.12

6.25

5.28

6.40

  

Large

Non

–0.02

–6.06

–0.06

–3.80

5.62

8.38

4.72

6.51

   

Uni

–0.18

–3.79

–0.62

–1.66

5.21

7.70

5.10

5.71

 

40%

Small

Non

–0.63

–4.45

–0.10

–3.82

5.05

7.12

5.34

6.86

   

Uni

–0.43

–2.59

–1.00

–1.96

5.44

6.73

5.05

6.80

  

Large

Non

–0.43

–10.68

0.32

–7.00

5.33

12.72

5.08

9.50

   

Uni

–0.67

–8.01

–0.52

–6.96

5.10

10.78

4.54

12.24

  1. Note. Unadjusted = multigroup IRT model with the common item parameters without adjustment to DIF, Adjusted = multigroup IRT model with the unique item parameters adjusting for DIF, NonDIF = NonDIF groups, DIF = DIF groups, Non = Nonuniform DIF, Uni = Uniform DIF. Bias and RMSE were computed for the PISA scale scores