Comparación de modelos de ganancia y crecimiento en rendimiento y sus efectos en las estimaciones del valor añadido de las escuelas

  1. NAVARRO ASENCIO, Enrique
  2. JIMÉNEZ GARCÍA, Eva
  3. GONZÁLEZ BARBERA, Coral
Book:
Actas XVIII Congreso Internacional de Investigación Educativa: interdisciplinariedad y transferencia (AIDIPE, 2017)
  1. María Esperanza Herrera García
  2. María José Rodríguez Conde
  3. Susana Olmos Migueláñez
  4. Fernando Martínez Abad
  5. Eva María Torrecilla Sánchez
  6. Juan Pablo Hernández Ramos
  7. Patricia Torrijos Fincias
  8. José Carlos Sánchez Prieto
  9. Adriana Gamazo García
  10. Francisco José García Peñalvo
  11. Antonio Miguel Seoane Pardo
  12. Valentina Zangrando
  13. Alicia García Holgado
  14. Felicidad García Sánchez
  15. Juan Cruz Benito

Publisher: Asociacion Interuniversitaria de Investigacion Pedagógica (AIDIPE)

ISBN: 978-84-697-4106-1

Year of publication: 2017

Pages: 1193-1204

Congress: Congreso Internacional de Investigación Educativa (AIDIPE) (18. 2017. Salamanca)

Type: Conference paper

Abstract

Change in academic achievement can be measured in two different ways, in the form of gain or growth. Gain is calculated using only two student performance scores (pretest-postest), while estimating growth requires more than two measures of educational attainment. This paper aims to verify how the use of measures of profit or growth affects the results of the evaluations. The sample consisted of a cohort of primary school students (893) grouped in 55 classrooms of different schools of the north of Spain, who were evaluated the mathematical competence by means of a standardized test at the end of the 4th, 5th and 6th grades. Three gain models (gross, residual and estimated), a fully nested growth model, and a status model with a single performance measure were estimated, the scores associated with the classrooms were calculated with each of the models and the outputs were correlated using parametric and non-parametric analysis. The results show high correlation between models with two and three scores, however very different results with the model that only uses a single performance score were found.