Table 2 Parameters and criteria in evaluating MDAS.

From: Developing and validating a short version of the multidimensional depression scale in asian adolescents sample using rasch analysis

Parameters

Indices & Criteria

1. Rating scale functioning – Monotonicity34:

Is the probability of selecting the higher categories in rating scale increases orderly with increasing trait level

35?

Number of observations in each response category: > 10

Average ability of individuals in each category: shows consistent growth

Outfits MNSQ of categories: within 2 to -2

Andrich threshold of each category: shows consistent growth

Andrich threshold differences of adjacent categories: within 1.4 to 5.0

2. Internal scale validity – Fit Statistics36:

Is the observed data compatible with the model37? Including38:

Person fit: how well the person response fits the expectations?

Item fit: how well the observed frequency in each item fits the expectations?

Person infit and outfit MNSQ: within 0.7 to 1.30

Item infit and outfit MNSQ: within 0.7 to 1.30

3. Construct validity—Local dependence39:

Is the response on one item independent of the influence from the responses on other items40?

Residual correlations of item pairs: ≤  ± .0.20 in Yen’s Q

4. Construct validity—Unidimensionality41:

Can the variances of all items be attributed to a single underlying trait42?

Percentage of variances explained by the measures: ≥ 50% in Principal Components Analysis of Standardized Residuals

Eigenvalues of the raw variances explained by the contrasts: ≤ 2.0 logits

5. Measurement invariance – Differential item functioning (DIF)* 43:

Do individuals from diverse groups exhibit consistent performance on an identical item set if they have similar ability level44?

DIF contrast: < 0.43

p-value of t-test: >0 .05

6. Construct validity—Item separation41:

Do the items within scale have sufficient difficulty division?

Item separation: ≥ 3.00

7. Item reliability

How replicable is the division of item difficulties in item separation?

Item reliability: ≥ .0.90

8. Discriminant validity—Person separation41:

Do the abilities of person have sufficient division in the scale?

Person separation: ≥ 2.00

9. Person reliability

How replicable is the division of person abilities in person separation?

Person reliability: ≥ .0.80

  1. *The DIF across genders was examined to explore if the shortened scale preserves the gender invariance as the original scale, despite the shorter scale length.
  2. The indices we drawn to examine the selected parameters were presented in italic.
Back to article page