What is the justification for adjusting past values, and is there any way to convey the increasing level of statistical uncertainty in the USHCN values,
like confidence intervals or error bars on charts?
I also think, rumbling around under the covers, there are fundamental misunderstandings about notions
like confidence interval which we statisticians just have not done a good job of educating about.
Not exact matches
TLDR; in statistics, a mean value is meaningless unless you know the
confidence interval for a given error probability (any poll saying that candidate Y will get X votes actually tells, in the fine print, something
like «we are 95 % sure that the candidate will receive at least X-error votes and at most X + error votes»)
It seems
like with a bit of math (
confidence intervals etc.) you could derive this info.
There are accepted methods by which one can manage uncertainty — statistical analysis,
confidence intervals, and the
like.
What I want to say is only that empirical evidence of the type that F+G 06 or any other of these analyses of climate sensitivity or related variables is information on the likelihood function (or equivalently conditional probability), and that this information alone can not provide any PDFs of
confidence intervals for the climate sensitivity or a functionally related parameter
like Y. To get a PDF or a
confidence interval, a prior must be assumed and plausible alternative priors give in this case significantly different results.
Ie,
Confidence intervals that are unrealistically small, you can call this an artifical reduction in variance if you
like.
I think that I've posted up on Foster and Stine on the calculation of «honest
confidence intervals» in stepwise regression (it's a phrase that I
like and use).
All three methods agreed that the effect of climate change was positive, making precipitation events
like this about 40 % more likely, with a provisional 2.5 - 97.5 %
confidence interval of 5 - 80 %.
Peter317 Fixed yours too: Moral of the story: on spans too short to have a significant
confidence interval you can put your start points and end points wherever you
like, and the trend line changes.
There's a reason science relies on things
like 95 %
confidence intervals instead of the ol' eyeball when looking at data.
By placing stations at what looks
like relatively even
intervals along major routes — as well as big box stores
like Target and Walmart — and by prioritizing multiple charge points at each location, and faster rates of charge (assuming your car can take it), this network will genuinely make a huge difference in the level of
confidence a driver might have as to whether they can find a charge.
The GMST trend for the last 40 or so years has converged to something
like a mean of +0.2 ºC / century, with considerable spread to the associated «95 %
confidence interval».
What I am doing is almost
like comparing the 99 %
confidence intervals of 2 variables, and saying that they are statistically significantly different, if there is no overlap.
Performed various statistics
like power, inter and intra subject calculations,
confidence interval and other point estimates.