Skip to main content

Table 2 Results of scale refinement process for each single-modality sensory subconstruct

From: Examining the latent structure and correlates of sensory reactivity in autism: a multi-site integrative data analysis by the autism sensory research consortium

 

Initial nitems

Final nitems

Final model fit

ρxxT

Scale retained?

TLIC2

RMSEAC2

SRMR

HYPER constructs

       

Auditory (Bifactor)

10

7

0.989

0.051

0.046

0.893/0.945

YES

 Unidimensional model

10

4

0.998

0.030

0.028

0.875/0.898

YESa

Visual

9

4

0.996

0.046

0.021

0.877/0.928

YES

Tactile (Bifactor)

18

13

0.985

0.043

0.047

0.916/0.946

YES

 Unidimensional model

18

4

0.978

0.053

0.038

0.718/0.713

YESa

Olfactory

3

3

0.017

0.751/0.840

YES

Gustatory

9

4

1.005

0.000

0.017

0.849/0.859

YES

Vestibular/proprioceptive [movement]

10

4

1.002

0.000

0.030

0.757/0.891

YES

HYPO constructs

       

Auditory

5

3

0.024

0.891/0.904

YES (Speech)

Visual

3

3

0.024

0.656/0.631

NO

Tactile/somatosensory

9

3

0.022

0.837/0.863

YES (Pain/Temperature)

Olfactory

2

2

NO

Gustatory

1

1

NO

SEEK constructs

       

Auditory

3

3

0.010

0.729/0.623

NO

Visual

12

6

1.002

0.000

0.027

0.834/0.859

YES

Tactile

10

4

1.011

0.000

0.039

0.746/0.727

YES

Olfactory

5

3

0.045

NO

Gustatory

5

3

0.044

NO

Oral tactile

5

4

0.985

0.088

0.026

0.848/0.938

YES

Vestibular/proprioceptive [movement]

10

4

0.993

0.042

0.029

0.831/0.818

YES

  1. HYPER, hyperreactivity; HYPO, hyporeactivity; SEEK, sensory seeking. Bolded values indicate instances wherein a scale did not meet a priori-specified criteria for psychometric adequacy (i.e., TLIC2 > 0.97, RMSEAC2 < 0.089, SRMR < 0.05 [or < 0.033 for 3-item composites], ρxx > 0.7, and ωT > 0.7). nitems, number of items on the scale; TLIC2, C2-statistic based Tucker-Lewis index; RMSEAC2, C2-statistic based root mean square error of approximation; SRMR, limited-information standardized root-mean-square residual; ρxx, marginal reliability; ωT, omega total (composite score reliability)
  2. aAs a bifactor model of these constructs demonstrated adequate fit and stronger reliability, unidimensional Auditory and Tactile HYPER item pools were retained for use in bifactor HYPER model only