That's the phrase I was looking for when I wrote my first response.
Standard Deviation is the best way to go. This is assuming that the
cheaters/lazy people are the minority. This will not work if they are
majority or even half of the end users. 

-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Carroll Kong
Sent: Friday, June 03, 2005 5:54 AM
To: The Hardware List
Subject: Re: [H] -OT- Logic question for programmers on the list

Eli Allen wrote:
> Part of my psychology class was on how to get better results on
surveys 
> so......
> 
> ----- Original Message -----
> 
>> Maybe add number 6, "No Comment"? LOL!
>>
>> Eli's idea sounds similar to a Stanton survey personality profiler.
>>
>>
>> Thane Sherrington wrote:
>>
>>> At 07:35 AM 02/06/2005, Eli Allen wrote:
>>>
>>>> answer doesn't always have the same meaning and allows for
detection 
>>>> of what needs to be thrown out by opposing questions)  Laziness is 
>>>> not the same thing as actively trying to protest.
>>>
>>>
>>>
>>> No it isn't, but I can't be sure which I'm dealing with.
>>>
>>> T

Maybe try a bit of both.  Try mixing up the questions a bit and since 
you are storing the data you can try different algorithms to throw out 
the best and worst cases.

Since you have the static data, you can play with the numbers and see 
which ones give the most similar results to each other.

So you can take the 80th percentile instead (if you had 10 
questionaries, you'd throw out 1 entry, one at the top or bottom.), and 
recalculate the average then.

Redo it at 90th percentile (throw out the top or bottom), redo it at 
70th and 60th.

Or you can find the standard deviation and throw out the top and bottom 
performers there and recalculate the average again.

Keep throwing out certain values and see if any of the results are 
similar.  Once you find the right "tweak", just use that from now on.



-- 

- Carroll Kong

Reply via email to