It has been 18 years since I took any statistics. I learned how to use SAS but have long since forgotten it and I'm sure there are all kinds of much more user friendly programs available today. I remember there was a formula that gave the confidence level if you knew the sample size and the standard deviation but couldn't find it if my life depended on it. Well maybe if... Anyway, if someone can provide me the formulas or point me to a web site, I'd appreciate it. Bob
One factor is if the 50 dealers you sampled represent the total 500 or so out there. Not only if 50 is enough, but if the proportions of big and small dealers is right and if other factors are represented. Once that is figured out, you just scale the average sales for the 50 up to the 500 and that's the answer.
50 out of 500 should be plenty good for a smaple size.
<but if the proportions of big and small dealers is right and if other factors are represented.>
that's really the key isn't it...
You really want a sampling of 30% of the stores in that 500 which do the most "through the door" business in the first place. The places that are best located and have the highest volume of traffic represented. If it's selling well at well located metro stores that get a lot of traffic those sales could easily negate anything that goes on in the bottom 1/2 of the 500 store list.
If your sampling doesn't pick up that diversity it could be biased to the lite side.
Well, it's been over 35 years since I've had any education in statistics. I did it, but don't remember it. Nevertheless, I don't think standard deviation, or any other statistical measure has anything to do with the price of KVHI. To me, it is simply supply and demand.
Your work is much appreciated, Bob. I will say I don't know how useful running a stastical program will be here given that you are surveying less than 10% of the stores carrying the product. The variables between regions (demographics, stage of economic recovery, average disposable income, etc.) seem to me to be too great unless one gets very detailed (e.g. breaks down the sampling by region, etc.).
A method you might consider is taking the numbers for the ~50 stores you have data for, omit the highest 10% and lowest 10% as outliers, then extrapolate sales based on the remaining data for 40 stores. Given the small ## of unit sales at issue (even if it is 3-5K, that's still small), omitting the highest-selling stores will probably skew the numbers a bit to the conservative side. It seems to me that is preferable to avoid unreasonable expectations. For instance, it seems your survey is focused on Florida, Georgia and a high-income area of southern California. I'm not sure the numbers generated would be representative, say, of sales in Utah or Arkansas.
Then again, what the heck do I know about statistics? I was an English major. :o)
Finally, more interesting to me than initial sales are follow-through orders. I fully expected there to be some pent-up demand for this product. Seeing follow-on orders from retailers is what would really make my day.