Crimes Against Statistical Inference: Forcing Teachers to be Accessories after the (Absence of) Fact.

Ridgway, R.; Ridgway, J.

Abstract

Reports on pupil performance form an important element in the efforts to improve the quality of education. Here, we examine the practicalities of making reliable judgements about changes in school performance over time. In a very large number of primary schools in England, too few pupils provide performance data to allow sensible conclusions to be drawn. In fact, the sample sizes are often not big enough to reliably detect the growth in attainment expected over a whole school year – let alone nuances about whether the performance by the same age group in adjacent years has changed. We provide links to simulations to illustrate this argument. Statistical inference is a difficult topic, and few teachers are able to draw conclusions from data that a professional statistician would endorse. We set out some heuristics developed by Wild and his co-workers (2011, in press) that can remedy this problem. We are critical of current practices and current plans to continue to perpetrate crimes against inference, and make suggestions about how this can be avoided. We highlight the strong negative impact that the poor statistical literacy embedded in government reporting requirements for schools can have on the lives of individuals, and on the reputation of schools.

Citation

Ridgway, R., & Ridgway, J. (online). Crimes Against Statistical Inference: Forcing Teachers to be Accessories after the (Absence of) Fact. Online educational research journal,

• Journal Article Type: Article
• Deposit Date: Feb 9, 2011
• Journal: Online Educational Research Journal
• Electronic ISSN: 2044-0294
• Publisher: School of Education, Durham University
• Peer Reviewed: Not Peer Reviewed
• Keywords: Curriculum, Educational Effectiveness, Leadership and Management, Teacher Training and Education
• Public URL: https://durham-repository.worktribe.com/output/1544033
• Publisher URL: http://www.oerj.org/View?action=viewPaper&paper=27