mean or group representation

[beginner_433]

I have 3 sets each contains 10 numbers. (Each number is a result of a computer program). I wanna take average ( mean) in each set, how ever in one set there are a few VERY good results which might change the mean in correctly. How can I get rid of Very bad/ very good results while calculating mean?

Thanks

[Mean Joe]

Hey you never said what this is for, eg do you want to compare the means?

My advice, based on what you've said, is that you could report more than just one baby statistic (the mean). Just the mean won't give a very complete description of the data. Add in standard deviation and range.

[beginner_433]

Hey you never said what this is for, eg do you want to compare the means?

My advice, based on what you've said, is that you could report more than just one baby statistic (the mean). Just the mean won't give a very complete description of the data. Add in standard deviation and range.

Hi,

Thanks for the reply. Values are classification accuracies in the form of percentages. e.g 67%, 65%, 71%,73%,74%,69%,98%
In above 98% will change the overall mean where as most of the numbers are around 60-75.

Can I do the following:
get the range= Max-min
then find c=rang/2
then define some boundaries: c+- constant value
and for each group I apply only numbers within that boundary?


boundary could be t in a way that [c-t c+t] has 80 percent of the numbers?


Please help,

thanks

Thanks

[Mean Joe]

Sure you could do that.

Maybe its good for what you're interested in. But if this was something like a stock price, I'd like to know that the price could jump 50% (from 60 to 90) every seven times (~15%).

If you would like to set a boundary as you suggest, then that is saying you're not interested in what those large "98%" values contribute, and that is often true, it depends on your interest. You're more interested in the reliable, routine values than the "freak values".