Table 1 Dimensions of validity assessed with hypotheses, evidence sources, and statistical criteria
Validation proposition: Use of the CPAS as a research measure assessing early psychosocial adversity as a child developmental risk factor among low-SES, urban Bangladeshi children aged 18–60 months | |||
---|---|---|---|
Validity dimension | A priori hypotheses | Sources of evidence | Statistical criteria |
Construct validity Scientific soundness of measured construct | (a) Cumulative early psychosocial stress, influenced by child and caregiver experiences, shapes human development (b) The scientific construct of psychosocial adversity has locally specific and embedded manifestations in Mirpur | Literature review | Not applicable |
Expert review | |||
Content validity Extent to which content captures construct | (a) Item content captures major themes in the conceptual model without extraneous content (b) Factor analysis will support a subscale structure corresponding to the conceptual model | Expert review | Significant item loading (e.g., ≥0.4) on primary factors, minimal cross-loading30 |
Cognitive pretesting | |||
EFAs | |||
Internal consistency Content cohesion | Subscales will show good internal consistency after final item selection | Cronbach’s α within subscales | Cronbach’s α ≥ 0.729 |
Test–retest and inter-rater reliability Stability of scores over time/raters | (a) Total scores will have acceptable test–retest reliability (b) Total scores will have acceptable inter-rater reliability, with lower reliability than for test–retest administrations due to layered variance related to rater and occasion | Retests over 2-week interval with same interviewer (test–retest) or different (inter-rater) | Average ICC ≥ 0.75 for test–retest (excellent), ≥0.60 for inter-rater (good)31 |
Convergent and discriminant validity Agreement with similar & distinctness from dissimilar measures | (a) Subscales scores will correlate with similar instruments, likely only moderately given non-identical constructs (b) Comparator instrument scores will correlate more strongly with associated CPAS subscale than with CPAS total score | Data from CPAS and comparator instruments | Pearson’s r with p < 0.05 |
Predictive validity Association with outcomes | Full-scale and subscale scores significantly predict future child cognitive performance, both in bivariate analyses and when controlling for other risks. | 48-month CPAS and 60-month WPPSI-IV scores | Pearson’s r with p < 0.05 |
Incremental validity Extent of novel value | (a) The CPAS will explore new domains of psychosocial risks while taking less time to administer than related instruments (b) It will show similar or better internal consistency | Age-matched data from CPAS and comparator measures | Cronbach’s α greater for CPAS greater than comparator measures |