SciFest beacons of hope

I've been a judge on the local young scientist exhibition/competition SciFest for the last five years and made comment on the experience, most recently last week. That post must evoke a picture of Uncle Bob peering crossly over his spectacles pointing out everything that's wrong with student presentations. But I'm not like that really and, truth to tell, I often forget to point out the failings of presentation as we roll up our sleeves and get down-and-dirty with the data. In the past, I've complained about the apparent lack of statistics in secondary schools. Another thing that we all note in the judging process is that certain schools - Avondale in Rathnew Co Wicklow, for a past example - seem to sweep the board of prizes on a regular basis. It must be something in the water the Science Teacher[s] who are either particularly inspiring or particularly competitive-by-proxy.

This year, I was chatting to two girls from Heywood Community School, which is really in the middle of nowhere. They had devised a project on the value of feedback in matters of learning. They split their class onto three cohorts who were assessed on their ability to carry out a learning task. Regardless of their results [a fib that would have them sent to prison by Immanuel Kant] each student was told either a) nothing (control) b) they had done good c) they were as thick as pig dribble. Two weeks later, the prospects were given a similar learning test and their two results compared. There was a hypothesis about whether praise or criticism would be better as a didactic tool and it was tested.
Suddenly in the middle of her breathless gabble, the young woman said ". . . so we did a t-test and then"
">!WHOA!< You did a t-test?".
"Yes, but we were worried about the variance, so we did an F-test and then we were confident that we should use Excel's t-test with unequal variances rather than making assumptions . . ."
Well, I was flabber-smacked! Here were two 16 y.o. girls from Ballygobogwards who had more statistical savvy than any of our Post Graduate students at The Institute. They admitted that they had sent their first attempts at data-analysis to a family friend who works for Pfizer in Cork.
Her: "Have you heard of Pfizer?"
Me "Yerrrrs, it's one of the biggest multinational Big Pharmas on the planet . . . oh I do apologise that sounds a bit patronizing"
Her: "I'm sorry if I sounded patronizing, I'm never sure what people know"
We thus had a bit of a pas-de-deux about who had out-patronised whom. It was such fun and I made a strong case to blag them the Math Prize in the compo.

Across the room, one of the other judges drew my attention to a Radon study. I was nothing loath to talk to them because we live with the stuff. And as it happened the girls [more girls - we needn't worry too much about upcoming women in science] come from just across our mountain at FCJ in Bunclody. What impressed me about their study was its power. They weren't doing radiation chemistry, which is hard to carry out in an under-resourced school science laboratory. But they were concerned that, although remediation for radon can be really effective, many of their neighbours were plug-ignorant about the lung-cancer inducing gas. That was a worry because, like us, Bunclody is built right on top of Mt Leinster granites. Granites are igneous rocks which have 'recently' boiled up from the Earth's interior laden with Uranium . . . which decays into Radon . . . which bleeds up through the floorboards . . . whence it enters your lungs . . . where it decays into Polonium . . . which sticks hydrostatically to the lung epithelium . . . where it sheds an alpha-particle . . . which mutates the DNA of the cells . . . which then develop into a carcinoma. The house that Jack built! Or rather the tumour built from a series of contingent events.  The girls made up an awareness questionnaire and delivered it to 40% of the households in Bunclody and <control> 40% of the homes in Clonegal, 5km North and built on a different geological formation. That's a lot of data in the SciFest context which is more likely to be the results of asking 5 girls and 5 boys how often they get their hair cut. I pointed out that 40% was peculiar; it was the absolute numbers that mattered rather than the proportion: our townland has 40 houses, 40% of that wouldn't be useful statistically.

Their data was essentially lots of Y or N answers from their questionnaire and they had a lot of elaborate statistics [again! that's two statistically savvy projects only 30 metres apart after 5 years of statistical desert - numeracy is in the wind] but I felt that elaborate statistics were not required . . . and I got the feeling that the girls had 'bought in' the formulas and the Z-statistics. Accordingly, I said that each question could probably be treated as a 2x2 contingency table, and chi-squared statistics applied. Chi-sq is a joy because it can be done, literally, on the back of an envelope [by teenagers even] or on the web and gives a really good handle on whether there is an association between two variables:

Spell Radon?	Bunclody	Clonegal	Row Tot
Y	120	30	150
N	80	40	120
Col Tot	200	70	270

Not real data: illustrative only. If the data were real then: the chi-square statistic is 6.17, and the p-value is 0.012 which says clearly that people who live with radon can spell radon whereas if it only affects those people over there, then you don't give a damn.  Notwithstanding my, not very well informed, opinion on the stats, those Bunclody girls were totally down on their project. I think it was sufficiently well designed and powered to inform government policy. Brilliant! . . . and they carried off the Intel SciFest 10th Anniversary Prize for innovation.