heartlogic-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

internal representation of scores (was Re: [Heartlogic-dev] wiki feedbac


From: Joshua N Pritikin
Subject: internal representation of scores (was Re: [Heartlogic-dev] wiki feedback)
Date: Thu, 19 May 2005 13:22:29 +0530

On Wed, 2005-05-18 at 18:51 -0500, William L. Jarrold wrote:
> On Tue, 17 May 2005, Joshua N Pritikin wrote:
> > You wrote:
> >> The below is Joshua's example and I find it confusing. E.g. what does
> >> "stats are positive" mean?
> >
> > In this context, positive means believable and negative means
> > unbelievable.  Internally, the Likert 1-5 scale is converted into a
> > [-1,1] range.  Hence, it won't matter internally if we start playing
> > with the ideas mentioned on your WhatScaleToHaveUsersRateOn page.
> > Everything boils down to [-1,1].
> 
> This is true if all of our rating schemes involve a fixed range.
> 
> But, if we ever go to a scheme where a score could be e.g. from pos 
> infinity and/or to negative infinity then this scheme won't work.
> 
> I like storing the raw rating numbers as it is an automatic debugging
> device, automatically keeps track of what ratings we allowed.
> 
> Normmalizing is importannt at the data analysis stage.
> 
> For naive users, the grid would be more sensible with raw data.  Many
> scientists/engineers don't even know what normalization is.  Plus the
> word is ambiguous - could be guassian, linear, etc.

I'm getting irritated by this thread.

Part of the problem is that I have omitted keys and legends on various
screens.  That's my fault.

However, arguing about the internal representation of data seems utterly
tangential.  If you want to change the internal representation then,
hey, go ahead.  The code is licensed under the GPL.  Patches welcome.
However, I would rather spend time on more important aspects of our
research.

I have added your objection to the wiki so that you can properly place
the blame on me if this design decision turns out to be wrong.

I hope this is the end of this thread.

-- 
If you are an American then support http://fairtax.org
 (Permanently replace 50,000+ pages of tax law with about 200 pages.)

Attachment: signature.asc
Description: This is a digitally signed message part


reply via email to

[Prev in Thread] Current Thread [Next in Thread]