Predictive Assessment of Reading

Reading

Rating Summary

Classification Accuracyfull bubble
GeneralizabilityBroad
Reliabilityfull bubble
Validityfull bubble
Disaggregated Reliability and Validity Datafull bubble
Efficiency
AdministrationIndividual
Administration & Scoring Time16 Minutes
Scoring KeyComputer Scored
Benchmarks / NormsYes
Cost Technology, Human Resources, and Accommodations for Special Needs Service and Support Purpose and Other Implementation Information Usage and Reporting

Annual cost per student depends on number of students as follows:

1-250 students for $7/student

251-500 students for $6.50/student

501-1000 students for $6.00/student

1001-2000 students for $5.50/student

           

Internet access is required for full use of product services.

Testers will require 1-4 hours of training.

Paraprofessionals  can administer the test.

Spanish supplement is available.

Red-e Set Grow 
Visit: www.OnlinePAR.net

Training manuals should provide all implementation information.

The Predictive Assessment of Reading (PAR) is a highly accurate, one-on-one, teacher administered, universal screening and diagnostic tool that can be given in 15 minutes or less to children in second semester Kindergarten through the beginning of fourth grade. A secure web generated report includes nine significant indicators of reading deficit problems, as well as reading proficiency. Included are an overall composite score that is highly predictive of a concurrent WJBR score, four standardized subtest scores, a diagnosed remediation priority code and a remediation intensity score, a highly accurate 3rd grade predicted WJBR score and a highly accurate 8th grade GM predicted score. PAR owes its high accuracy to an algorithm derived using a unique combination of four standardized subtest scores for Picture Vocabulary, Phonemic Awareness, Rapid Naming, and Letter-Word calling, each of which is statistically significant as independent contributor to the total prediction. Long term longitudinal tracking of control groups provided the basis for the multi-year predictions. Extensive analysis revealed that addition of subtests such as nonword reading, visual perception, memory, or comprehension did not yield any significantly stronger predictions. Compared to all other currently used predictive devices or algorithms, PAR is unique in its inclusion of picture vocabulary as one of its subtests. PAR can be differentiated by its unique ability to provide the teacher with a diagnosis of the single or double deficit that must be corrected before the child will be able to advance. Based upon this global diagnostic profile, PAR provides the teacher with a strategy and a starter set of 20 minute scripted remediation lesson plans.

Raw PAR scores are generated for the addition of correct student responses, or time in seconds for each subtest. Six raw scores are entered into a secure website scoring engine and a proprietary algorithm, based on national norms, which generates the standardized subtest scores. The composite score is calculated using another algorithm that combines the four standardized subtest scores along with grade and month of school year weighting.

Individually administered.

 

Classification Accuracy

Classification Accuracy in Predicting Standard Score on the Woodcock-Johnson III Broad Reading Composite
  n=500
False Positive Rate 12.47%
False Negative Rate 9.57%
Sensitivity 90.43%
Specificity 87.53%
Positive Predictive Power 68.42%
Negative Predictive Power 96.84%
Overall Classification Rate 88.2%
AUC (ROC) 0.960
Base Rate 23.0%
At 90% Sensitivity, Specificity equals 87.53%
At 80% Sensitivity, Specificity equals 93.77%
At 70% Sensitivity, Specificity equals 96.63%

 

Generalizability

Description of study sample:

  • Number of States: 6
  • Race/Ethnicity:
    • 58.0% White, Non-Hispanic
    • 19.8% Black, Non-Hispanic
    • 20.0% Hispanic

Cross Validation Study Description of study sample:

  • Number of States: Florida
  • Size: 200
  • Gender:
    • 52%  Male
    • 48%  Female
  • SES: 38% eligible for free or reduced-price lunch
  • Race/Ethnicity:
    • 68% White, Non-Hispanic
    • 10% Black, Non-Hispanic
    • 15% Hispanic
    • <1% American Indian/Alaska Native
    • 2% Asian/Pacific Islander
    • 5% Other
  • Language proficiency: 5.5% of students had a primary language other than English

Reliability

Type of Reliability Age or Grade n
(range)
Coefficient
Range Median
Cronbach’s alpha for each of the three subtests K, 1, 2, 3 500 0.90, 0.92, 0.93 0.92
Rapid Naming Test-retest   Later group N=50 0.92  
Alternate forms 1 Later group N=50 0.90  

 

Validity

Type of Validity Age or Grade Test or Criterion n Coefficient Range
Concurrent K, 1, 2, 3 WJ-III Broad Reading 500 0.91
Predictive 3 Florida Comprehensive Achievement Test 703 0.77

 

Disaggregated Reliability, Validity, and Classification Data for Diverse Populations

Classification Accuracy in Predicting the Proficiency on the Woodcock-Johnson III Broad Reading Composite
  African-American
n = 99
Hispanic-Latino
n = 100
Caucasian
n=290
False Positive Rate 0.23 0.06 0.14
False Negative Rate 0.19 0.09 0.12
Sensitivity 0.81 0.91 0.88
Specificity 0.77 0.94 0.86
Positive Predictive Power 0.72 0.97 0.38
Negative Predictive Power 0.85 0.83 0.99
Overall Classification Rate 0.79 0.92 0.87
AUC (ROC) 0.90 0.97 0.97
Base Rate 0.42 0.35 0.35

Disaggregated Reliability

Type of Reliability n Coefficient
Test-retest African American 48 0.90
Test-retest Hispanic-Latino 31 0.94
Test-retest Caucasian 108 0.83

Disaggregated Validity

Type of Validity Test or Criterion n Coefficient
Concurrent Validity
African American
WJ-III Broad Reading 99 0.86
Concurrent Validity
Hispanic Latino
WJ-III Broad Reading 100 0.92
Concurrent Validity
Caucasian
WJ-III Broad Reading 290 0.92
Predictive Validity
African America
FCAT 48 0.77
Predictive Validity
Hispanic Latino
FCAT 31 0.78
Predictive Validity
Caucasian
FCAT 108 0.71