| |
 |
|
|
Science Forum Index » Statistics - Math Forum » Positive Correlation and User Surveys
Page 1 of 1
|
| Author |
Message |
| Graham Ashe |
 Posted: Fri Mar 09, 2007 4:26 am |
|
|
|
Guest
|
| I have a set of data from a computer program on the economic value of chess (checkmate) positions. I need to demonstrate that there is a positive correlation between the computer's evaluations and the assessment of human problem composers. I'm thinking of showing pairs of checkmate positions selected from the data set to 30 or so composers and asking them to select the one which they think is more economical. Perhaps 5 pairs of positions that the computer program has evaluated with an increasing discrepancy between each pair (e.g. 0.1 between the first pair, 0.2 between the second etc.). What kind of statistical test could I perform on the data gathered from this to show the correlation I need? Thanks in advance. |
|
|
| Back to top |
|
| Graham Ashe |
| Posted: Sun Mar 11, 2007 3:27 am |
|
|
|
Guest
|
| One more thing. I'm not sure if using a 4-point Likert scale would be suitable for something like this. I would much rather just let the respondents decide which of the two positions is more economical and base the correlation on the percentage of positive or negative responses. What do you think? |
|
|
| Back to top |
|
| Graham Ashe |
| Posted: Sun Mar 11, 2007 8:36 pm |
|
|
|
Guest
|
Quote: I would not call it a Likert scale if it just happens
to have integer points.
But I would see if my raters
could give a measured judgment, rather than Yes/No,
just in case it helps. You can always collapse
ratings if
the raters do not give useful variation.
I suppose 'strongly agree', 'agree', 'disagree' and 'strongly disagree' options could be associated with a rating of 4, 3, 2 and 1 but it would be hard to associate that with the computer's evaluations that are usually between 0 and 1 and precise up to 5 decimal places. Also, the problem with a Likert scale is that I would have ask users questions based on a predetermined decision (e.g. "Position A is more economical than B. What do you think?") rather than letting them decide on their own (e.g. "Which position, A or B, is more economical?"). I think the second option is a better gauge of their responses.
Quote: By the way -- I wonder about your data collection.
The data? The question? You might ask more
than one question. "Economical" is an interesting
term,
but I can imagine more than one way to interpret it.
"Obvious" and "elegant" are words I remember from my
own
chess-playing days. Do chess players have a single
notion of economy?
Well the computer takes into account various aspects of economy. Giving a detailed description of what economy means would pretty much ensure that players will come to the same conclusion as the computer, which makes the survey pointless. Since my users will probably be problem composers, they should understand what economy means and the result would better indicate if the computer's assessment correlates positively with human perception.
Quote: Hey, if it is a *checkmate* position, how could it
be
any better, as positions go?
Essentially, it's the style in which the checkmate was performed. For example, using more pieces than necessary (given the position) is poor economy. |
|
|
| Back to top |
|
| Richard Ulrich |
| Posted: Sun Mar 11, 2007 8:45 pm |
|
|
|
Guest
|
On Sun, 11 Mar 2007 09:27:39 EDT, Graham Ashe
<knight_armour@yahoo.com> wrote:
Quote: One more thing. I'm not sure if using a 4-point Likert scale would be suitable for something like this. I would much rather just let the respondents decide which of the two positions is more economical and base the correlation on the percentage of positive or negative responses. What do you think?
I would not call it a Likert scale if it just happens
to have integer points. But I would see if my raters
could give a measured judgment, rather than Yes/No,
just in case it helps. You can always collapse ratings if
the raters do not give useful variation.
By the way -- I wonder about your data collection.
The data? The question? You might ask more
than one question. "Economical" is an interesting term,
but I can imagine more than one way to interpret it.
"Obvious" and "elegant" are words I remember from my own
chess-playing days. Do chess players have a single
notion of economy?
Hey, if it is a *checkmate* position, how could it be
any better, as positions go?
--
Rich Ulrich, wpilib@pitt.edu
http://www.pitt.edu/~wpilib/index.html |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Sun Oct 12, 2008 2:58 pm
|
|