Veljko’s comment > Likes and Comments

I am rushing, so can't address each point as it fully deserves.

But yes, there is a strong trade-off between how refined a scale is, and how much noise there is in the scale definition.

If we had a 'thumbs up' or 'thumbs down' scale, we would all have a pretty good idea. Up=good, down=bad. You put it on 5 points, you put in ambiguity. You use 10 points, the ambiguity intensifies.

In a previous life I was a statistician and I did a bit of work on survey design. It is a well documented effect.

I am not saying 10-points would not be desirable. I am just telling you it's not as simple as it may seem.

Oh regarding editing. No, I don't think someone would seriously modify their rating because of fond choice. But editor-introduced spelling errors could make a substantial difference, for example.

Hi Veljko,

Firstly, I believe this statistical "ambiguity" you refer to is actually called subjectivity in this case. So yes, you are right, in a ten star system the ratings are a more accurate reflection of the subjective whole.

Secondly, you wrote: "I am not saying 10-points would not be desirable. I am just telling you it's not as simple as it may seem". Hmmm... I must admit, this whole thing does seem exceedingly simple to me. People read a book and score it out of ten. The scores are amassed and we get an average score. Anything else?

Thanks

It seems simple to you - but it is not. Which is all I am trying to say.

Regardless. No, what I am talking about is not "subjectivity" - it's noise. As in any other process, we have some measurement error - and the purpose is to separate the signal - the true opinion - from the noise - which arises mainly from:
1) not being able to score accurately (and that is improved by moving to a 10-point scale) and
2) not understanding the scale (and that is made worse by moving to a 10-point scale).

For more information on point number 2, you can refer to a vast literature on psychology and survey design (neither of which I am an expert on, but know something about). If you are curious, you can start from a fairly simple, if dated, exposition my Miller (1956): http://www.musanim.com/miller1956/

His main argument is that the ideal number of points is seven - but the main message is what matters - too many points create noise.

If disagreement offends you guys, I apologize for commenting on your thread and will refrain from doing so further. I was under the impression that you were seeking an opinion - and I do stick to the statement that a more refined scale does not necessarily produce a more accurate score, especially if not accompanied by very clear definitions.

Sorry for the typos. It's been a long day at work...