Sunday, 8 January 2017

How NASA doctor surface temperature readings

How NASA doctor surface temperature readings
Comment by 'Pete Statistician'
So let me understand. I do a bit of statistics, so here goes. NASA takes a 100 year period of data, but the first bunch of years has a lot of "less factual" data in it, because there were not enough reliable collectors at various points around the world. But it is necessary for NASA to have a 100 year period, because a 30 year period is just not statistically significant for doing an extrapolation, yes? So NASA takes a calculation of the first 70 years, then performs "adjustments" on the data, which just happen to be uniformly downwards, and then puts these adjusted data points in its very convincing graph. So some non-statisticians (ie 99% of the people looking at the graph) can stare at the graph, see the hockey stick with the up on the right end, and draw non-statistically based conclusions? Well, just 2 points: 1) extrapolations are extremely risky - especially if you have not controlled for outside lurking variables, or other factors 2) a really short hockey stick (only 30 years) would not look that convincing, would it - it would also allow reasonable people to dispute the extrapolation by pointing to the flatter shape of the past 3) oh gee, let's add ethos by taking into account the first 70 years, but applying a meaningful "downward adjustment" to many of the data points, which just happens to help support the conclusion we want So, if you use the unadjusted data, you get much less of the slowly climbing result NASA wanted, and your alpha is a heck of a lot higher - ie, there is no meaningful relationship between year and temperature. Oh well, I guess when your funding depends on Obama the other rocket scientists, you do what you gotta do. I mean, we all have to survive, yes?

No comments:

Post a Comment