 |
 |
Interesting points, Alan! Precipitation probabilities always seem to be vague as to whether they refer to a time-space POINT, or a time-space LINE. e.g. this one says 9am 10%; 10am <5%. Does this mean 10% prob of raining at exactly 9am, or 10% prob of raining some time between 9am and 10am, or even 10% prob of raining THROUGHOUT THE TIME between 9am and 10am. Similarly on the space dimension. If these probs are for 'York' say, is the event: "Rain somewhere in York", or "Rain at a particular point defined e.g. by Google to be 'York'", or even "Rain throughout York"? This gives 3x3=9 possible interpretations for the event "Rain tomorrow at York". Even discounting the 'throughout' interpretations, but that still leaves 2x2=4. Can the real met office please stand up (with umbrella if need be), and advise us on this? The world needs to know! JOHN BIBBY On 2 September 2014 11:28, [log in to unmask] < [log in to unmask]> wrote: > The UK Met Office now supplies forecasts on the web with an hourly > probability > of precipitation. As the National Curriculum for schools now includes > teaching > probability, how would you suggest handling these values? For today > (below), the > hourly probabilities are obviously serially correlated. Any suggestions > for how > to > combine the p-values to get an estimate for the day? As a school exercise, > should I > take a coat or umbrella when I go out? > > 0900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 > Precipitation probability > 10% <5% <5% <5% <5% <5% <5% <5% <5% <5% <5% <5% <5% <5% <5% > > The "weather" is shown as "overcast" all day. > > Writing this at 1100, I can report the p value for 1000 and 1100 were > updated to > 10% > *after* 1000 and that at 1105 there is light rain from flat clouds. > > We find it strange that forecasts for this area are so erratic and > unreliable > given that > we are (a) roughly 40 miles NE of the Met Office in Exeter and (b) roughly > 6 > miles SE of > a Met Office rainfall radar station. > > My own thought is that the serial correlation has been discounted, and the > actual meaning was > that it was quite likely to rain today, but they could not say exactly > when. So > looking at <5% > across the day is not too helpful if you are going out for several hours. > > At 11:15 the rain has got heavier. > > Allan > > You may leave the list at any time by sending the command > > SIGNOFF allstat > > to [log in to unmask], leaving the subject line blank. > You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.