# Are national averages statisically significant for very small cohorts?

Discussion in 'Headteachers' started by meister, Jan 12, 2019.

1. ### meisterNew commenter

Our LA adviser insists on relating all data to 'National averages' despite us have very small cohorts, often less than 10. This is not proving helpful but need evidence (papers/reports/maths, etc.) to shake his mindset and prove it is statistically suspect to compare the national cohorts with ours.

He is data driven and obsessed with 'National Averages' and 'progress from starting points' - aarrrgghh!

2. ### strawbsEstablished commenter

Flanks and digoryvenn like this.
3. ### cornflakeSenior commenter

Also see Making Data Work publication 2018!

digoryvenn likes this.

5. ### cornflakeSenior commenter

School and trust leaders should not assume that group analysis of attainment will be more informative than whole class analysis. The smaller the groups being compared, the more likely that any differences observed are simply statistical noise. The dangers of over-interpreting analysis of small subgroups is very real: for example, in a one form entry primary school with 30 percent pupil premium pupils, you would need a 40 percentage point difference in pass rates between groups to report statistically meaningful attainment gaps.

I assume, since you are talking about comparing <10 children, with the National - the same applies...

digoryvenn likes this.
6. ### abacus1982Established commenter

Look at your Inspection Dashboard Summary Report. This will tell you whether any data is statistically significant. To give you context I have some data which was +/- 14% points from national but it states it is not statistically significant. This is because of the cohort size. It would be statistically significant if we were a larger school though.