@WingnutSOS @DenizenOfAisleOfShame
The issue with AI is that it is trained to give you a layman’s explanation for complex data questions without looking into the additional factors.
The ‘bowl of soup’ analogy is quite famous and is used to explain to early stats students how to use economical polling.
However where this analogy fails in this case is that it assumes that each nation’s education system is homogenous, where in reality each nation is deeply segregated by socioeconomic status, region, funding and ethnicity/culture.
This is where earlier I pointed out that rural schools in Colombia and Buckinghamshire aren’t the same, a closer example may be that a deprived inner city school in London which has a high rate of Chinese and Indian students (who are the demographics that have the highest educational achievement in the uk) would be deemed the same as a deprived school in Dundee attended mostly by white working class kids (who are the lowest achieving group in the UK).
So in basic terms the ‘soup’ hasn’t been stirred in the same way because the ingredients are different.
This paired with sampling only 0.8% of a huge population means sampling is far more vulnerable to unmeasured local variations than sampling 13.8% of a smaller, more uniform population.
In my last post I mentioned non-participation and selection bias, with only 0.8% responding Even a small drop introduces a huge level of selection bias favouring higher achieving schools.
Know I’ve got a habit of droning on about data but there are several other factors why the methodology can’t really give a like for like comparison, don’t mind typing it up if you’re really interested?
As I say though it isn’t useless, as an indicator of what we achieved compared to past performance it is actually very good.
Clearly we have declined and it shows that action needs to be taken.
That is what we should be taking away from the data, but as you have pointed out due to political agendas many people just want to point out how we are better or worse than our neighbour when really the benchmark we should be looking at is our own past performance.