Computational Application for Early Detection of Dyslexia in Preschool Children

1
Computational Application for Early Detection of
Dyslexia in Pre-school Children

• Background
• Aim of overall project
• Aim of current phase
• Phonological Deficit Hypothesis

2

• The most commonly used tests in Ireland at
  present
• The Dyslexia Screening Test (6.6 to 16.5 years)
• The Bangor Dyslexia Test (test not normed)

• Genetic Factor (Rack, 1999)

3
Developing the test

• Early/Intermediate/Late Consonants (Smit Hand,
  1997)
• 38 test items based on 19 phonemes
• Selecting words and images
• Pre-pilot the test initial idea of validity and
  reliability
• Technical aspects

4
(No Transcript)
5
Test Administration

• 8 pre-schools
• 104 children ( 20 at pre-pilot stage)
• Age range 33 to 69 months
• Locations rural Wicklow to urban City Centre
• 20 questions per child
• Mean score 11.89 Mode 11

6
Test Analysis
7
Point-Biserial Test

• An item analysis test
• Establishes a correlation between a right/wrong
  score a child receives in a given item and the
  total score that a child gets in the overall test
• Ranges from 1 to 1
• Varma Point Biserial of at least .15

8
Test Analysis contd.

• Low Point-Biserial Scores Removed

9
Cronbachs a

• A split-half measure of the reliability of the
  test items

• Cronbach's Alpha

.830
17

• Number of Items

10
Early/Intermediate/Late Consonants

• Developmental Sequence as recognised by SmitHand
  (1997)
• Early Consonants Mean 32
• Intermediate Consonants Mean 33.8
• Late Consonants Mean 32.58
• No significant difference between groups

11
Location

• Significant variation in scores
• Parental interest?
• Phonological training?
• Age variation?

12
Age-Sensitive Test

•  
•  
• Mean Score Standard Deviation
• Group 1 8.05
  3.268
• Group 2 11.90
  3.919
• Group 3 12.35
  4.499
• Group 4 13.30
  3.975
• Group 5 13.54
  3.833

13
Achievement

• 30 good items in terms of reliability
• Discriminates between children on their success
  at a phoneme-based test
• Age sensitive
• Correct pitch format of test

14
Futher Development

• Further pilot-test
• More care in getting a representative sample
• Try on a group who are known to have dyslexia
• Further statistical analysis
• Z-scores
• Multiple Regression