Testing the Structure of HELP 3-6 Assessment among Head Start Children using Confirmatory Factor Analysis

• In Event: Poster Session 13
  In Poster Session: PS 13 Topic - Methods, History, Theory

Abstract

The primary goal of the present study was to evaluate the factor structure of the HELP® 3-6 with data collected from children attending Head Start using confirmatory factor analysis. The HELP® 3-6 is an early childhood, curriculum-based, assessment tool designed for early childhood professionals to measure skill development among preschool-aged children. HELP® 3-6 consists of six domains (Cognitive, Language Gross Motor, Fine Motor, Social, Self Help) comprised of 47 strands (all of which are listed in Table 1) and one stand-alone strand (Self-Regulation) that does not fall under a separate domain. Strands are considered solitary constructs in which age-graded indicators progress from 36 months through 72 months. There are 480 age-graded indicators in the entire assessment that are marked as “fully,” “partially,” or “not having” the skill/behavior. Age-graded indicators for a given strand were designed to align with increasing maturation. Based on these, developmental age scores are created for each domain.

Data were collected as part of an ongoing evaluation of Head Start students in southwestern portion of the U.S. One thousand, one hundred, ninety-five individual preschool-age Head Start students were included in the dataset. Children were initially assessed with the HELP® 3-6 in October through November, a couple months after the academic year began. Students were 47.751 months old on average (SD = 7.164) and ranged from 36 months to 66 months. The dataset analyzed included score information at domain, strand, and indicator levels. The dataset also included student date of birth, age at time of assessment, and date of assessment.

We analyzed the degree to which HELP® 3-6 strand scores load to latent factors representing domains as indicated by the assessment. CFA models were calculated and fit using M-plus. Six CFAs were conducted, one for each domain. An omnibus CFA was not conducted because developmental age scores would be undistinguishable across assumed latent factors in a single test. It was expected that each strand associated with a given domain would yield significant load estimates with the latent factor if they represented the assumed domain. Estimates presented in Table 1 indicate the degree to which each of the strands contributes to the latent factor. Higher (closer to 1) coefficients indicate more contribution to the latent factor compared to lower (closer to 0) coefficients. P-values less than 0.001 for this size sample are a conservative indicator that the coefficient is contributing significantly to the latent factor.

As shown in Table 1, CFA results indicated each of the six models fit the data well. All of the strands were significantly associated with their respective domain, as specified by the assessment protocol, and most strands yielded strong coefficients with their respective domain. Three strands in the Social domain yielded coefficients below 0.50, indicating lower influence on that domain score. These findings indicate good structure for most domains of the HELP® 3-6 assessment. Lower coefficient values for the Social domain and overall findings will be discussed.

Authors

• Louis Manfra, University of Missouri - Columbia
  Presenting Author

• James A. Larsen, University of Tennessee - Knoxville
  Non-Presenting Author

• Danielle Turley, California State University - Stanislaus
  Non-Presenting Author