> My (many decades old) memory of statistics is that a standard deviation > assumes a Normal distribution. I suspect that the distribution of how many > light are left on in a building is a long way away from normal.
Well you're wrong. The graph is smooth and continuous. If you want to get ultra-technical, I use the term "normally distributed" to be simpler and enough people are going to misconstrue that, but since it's not symetrical it's actually a Weibull Distribution (http://people.stern.nyu.edu/adamodar/New_Home_Page/StatFile/statdistns.htm , about half way down) since it's skewed left. It's skewed left since you can't use less than zero electricity (insignificant amounts of the population are net generators), but otherwise follows pretty much perfectly smooth and expected logarithmic curvatures, especially to the right on the high end. Here's a PDF: http://www.lowcarbonlivingcrc.com.au/sites/all/files/publications_file_attachments/statistical_analysis_of_driving_factors_of_residential_energy_demand_-_final.pdf Look at basically any of the graphs starting at page 9. They've broken it down 10 different ways, they all follow the same shape. If you hate PDFs, here is a direct link to the relevant graph: https://i.imgur.com/Uk5bfxo.png The point is that it's smooth and tapers, and how exceptionally rare it is for someone to be at 10x the rate of anyone else. The average on this graph is around 15kwh/day, look how far to the right 150wh/day is. Even for this massive sample size (3300 households) it's an immeasurably small amount of the population. Even the amount to the right of 4x the average (60kwh/day) is less than 1%. > We have this same debate every time someone mentions that the "average" > person drives 35 miles per day; so an EV with a 50-mile range is fine. But > everyone jumps in to say they don't know *anyone* who drives 35 miles/day -- > they all drive 100+miles/day, or virtually no miles most days. Again, that is exactly the purpose of not just looking at a blind average but to also look at the standard deviation (or lambda, or whatever is relevant for the distribution math). The thing you're confused about is already included in what I was saying. Look at the graph I linked above. That is the raw data itself, not just average it's the actual data. If what you're saying is true there would be a double-peak, one low one high. We do not see that. We see nice, smooth, distribution as we head to the right. I think that's pretty conclusive. _______________________________________________ UNSUBSCRIBE: http://www.evdl.org/help/index.html#usub http://lists.evdl.org/listinfo.cgi/ev-evdl.org Please discuss EV drag racing at NEDRA (http://groups.yahoo.com/group/NEDRA)
