I was speaking to my stat teacher today, who said there was some Z*s/(n)^(1/2) formula that ends up giving you the minimum n for reliable statistics. Any idea?

 

Hi Jason:

___Although I haven’t had a stat class in over 15 years and my last semester long one was well before that, a reliable statistic at a given 95, 97.5, 99% etc. confidence with a given sample size relative to a given - total population is a first year type calculation. I have seen it in the past but I do not know what it is anymore. A google on 95% confidence, sample size, and total population should lead you to what you are looking for. If you do not find it, pose the question to your stat teacher like this:

What is the confidence level of a given Prius II owner to achieve 47.9 mpg with a sample size of 154 vs. a total population of 88,416 (total 04 and 05 sales through March 2005)?

___Make sure you mention that the FE posted in terms of tank over tank is by a population that may or may not be representative of the general population as a whole given the caveats of hypermiler tank uploads vs. non-hypermiler tank uploads. See below. She may know of a way to remove some of our lmpg’s as well as the low ***** from the calculations to give a new member/reader a real sense of what their average may or may not be from a given hybrid.

___The problem as I see it is that many of the GH members - tank over tanks are not complete. You have a general idea as to what an average FE may be for a given automobile but with 25% or more of the tanks posted coming from hypermilers, it would tend to scare off the lesser FE receiving members from posting lesser tanks. This leads to a higher average then what the general population may in fact receive that cannot be adjusted for with statistics. Maybe a psychologist would need to be contacted to remedy the skew us hypermilers instill into the overall data

___Just last night I removed the Overrides from the Hypermiler list because of anomalies I was beginning to see. It is very easy to upload “I get 48 mpg from such and such a vehicle” vs. posting actual tanks from receipts and the odometer and the ease at which to stretch the truth a bit just once is a lot easier then once, twice, four times a month if you know what I mean. IIRC, you do not use the Overrides in a particular automobile type or overall stat’s, right?

___Finally, I do wish you would increase the digits in the FE calculations just a bit although I dread re-editing all of the tanks posted to date Every gas station I have filled up at in the last 4 years has a xx.xxx gallons or down to (1/1000) of a gallon pumped shown. This being the actual is probably mis-leading of course but I always post it in the notes … The odometer or at least Trip A/B shows down to a tenth. Your Odometer entry w/ no digits to the right of the decimal point is a bit odd imho.

PS: Your new edit mode stinks. It produces edits in a quasi - hypertext mode w/ huge spacing, bolded text in some areas and not in others, weird <font> type code and such … Editing the Hypermile list or even this post is a huge pain in the @$$ because of this Is there a way to change the Edit mode in my CP somehow? It started just the other day …

___Good Luck

___Wayne R. Gerdes
___Waynegerdes@earthlink.net

 

I honestly don't think my teacher knows how to figure that out This entire year, we've never dealt with population size unless it was proportions - this is means.

Also, the data has little skew. If you'll look at the means and medians, they're essentially the same.

The database will now calculate everything with unrounded numbers. It just spits values out to tenths. I just made this change (it was a bug) a few days ago, so if you wan't accuracy you'll have to edit tanks. Sorry

 

[offtopic] You can edit that here. Not sure what you mean about the problems, though. I don't have any.