Page 1 of 1

Correlations and statistics for my webteases

Posted: Tue May 24, 2011 3:32 am
by cactusman
I have run some correlations (using Excel) on my web teases (all of my 84 teases, and no one else's :-)). Some points of interest to you may appear.

Some positive correlations:

The strongest positive correlation (aside from the unilluminating one between date and tease number) is between number of views and number of people who rated the tease, which is a correlation of 0.64. This just says that when more people view a tease, then more of them rate it. I also found that on the average only about 4% of the views generated a rating by someone. :-/

The second strongest positive correlation, which I have heard others (Evals?) hypothesize before, is a correlation between the teases number of pages and the rating of the tease, which is a correlation of 0.56. This might be because people like long teases, or because more people who don't like a tease give up early.

The correlation between ratings and number of pages may also explain the correlation of 0.52 between number of pages and number of views. I hypothesize that people tend to view highly rated teases more (although the correlation between views and rating itself is only 0.24). Then very slowly, as more people rate them, they tend to go down in rating if they aren't so good. Or it may be just that my most interesting work often turns into fairly long teases. I don't know. :-D

A somewhat pleasing positive correlation of 0.43 is between tease number (all teases have a sequence number) and rating. This might mean that my work is getting better over time :-), or simply that not enough people have rated the more recent teases to bring down their ratings yet. :-/

The strongest negative correlations are the ones are between number of raters and tease (sequence) number (-0.69) and number of raters and date (-0.70). But these just mean that older teases get rated more often.

I've written 84 webteases that are published so far.

Some averages: The average rating of all my teases is 4.233 (median 4.3). Anyone have a better average or median than that?
The average number of raters of all my teases is 361 (median 261). I thank those of you who like and rate my teases for their feedback. The average number of views of my teases is 8440 (median 7236).
And for what it's worth the average length of my teases is 92 pages (median 54).

You can easily find which of my teases have the highest ratings or the most views. But it's interesting that the tease with the most ratings is neither the highest rated or the one with the most views, but:

Total Devotion: http://www.milovana.com/webteases/showtease.php?id=52

which is only has the 5th highest number of views.

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 7:07 am
by stilev
lol not to nerd out on you, but would those numbers be the R^2 values? if they are you might want to run a regession analysis on the them to determine the p-values on the them, any values that are under a value of .05 are considered to be 95% accurate and is the normal the accepted value. you can have a high r^2 value, but without a high p-value they dont mean alot otherwise. the p-value says how strongly supported the r^2 value is. a r^2 value merely means what percent of the data can be accounted for by the y-variable.

sorry to nerd out for a sec just trying to help lol, it is rather interesting i think

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 7:35 am
by doti
a correlation between the teases number of pages and the rating of the tease
I thin most of us have a stupid habit of rating the writer, not the tease. Sus more pages => more efforts => higher grade.
0.43 is between tease number [...] and rating
:bow:

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 10:41 am
by cactusman
doti wrote: I thin most of us have a stupid habit of rating the writer, not the tease. Sus more pages => more efforts => higher grade.
I hope that most rate the writing, not the writer or the number of pages. But I don't think the correlation of 0.43 is very high in any case.

I don't actually know how to compute these p values. All these "correlations" are what is produced by MS Excel by using the data analysis "correlation" function. Sorry if I don't know more...

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 10:46 am
by les



*chuckles*
As to how you keep your name and teases in our minds.

well done.

I admire the thought and effort.
so very unlike the spammers.

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 10:57 am
by cactusman
les wrote:
*chuckles*
As to how you keep your name and teases in our minds.

well done.

I admire the thought and effort.
so very unlike the spammers.
Thanks Les. Sorry if anyone considers this spam :-D

I was also going to add that I was interested in the question of how the ratings were affected by whether the tease was original or not (as I have redone several classics and illustrated stories by other people). It seems that the original ones are more highly rated, as their average is about 4.38 (median 4.5). This is much better than the total of all mine, which have an average of 4.23 (and a median of 4.3) However, my original teases are also significantly longer on the average than my total output, with an average of 148 pages (median 120). So the correlation between length and rating might explain some of their higher ratings. Nevertheless, I am generally no longer re-illustrating stories from others and no longer illustrating letters from others, because the original ones seem more highly appreciated (and because I've already illustrated my personal favorites).

Re: Correlations and statistics for my webteases

Posted: Tue May 24, 2011 11:49 am
by les
cactusman wrote:
les wrote:
*chuckles*
As to how you keep your name and teases in our minds.

well done.

I admire the thought and effort.
so very unlike the spammers.
Thanks Les. Sorry if anyone considers this spam :-D


I exactly meant its not spam,
its everything spam ain't


cactusman wrote:I was also going to add that I was interested in the question of how the ratings were affected by whether the tease was original or not (as I have redone several classics and illustrated stories by other people). It seems that the original ones are more highly rated, as their average is about 4.38 (median 4.5). This is much better than the total of all mine, which have an average of 4.23 (and a median of 4.3) However, my original teases are also significantly longer on the average than my total output, with an average of 148 pages (median 120). So the correlation between length and rating might explain some of their higher ratings. Nevertheless, I am generally no longer re-illustrating stories from others and no longer illustrating letters from others, because the original ones seem more highly appreciated (and because I've already illustrated my personal
and then he adds more figures
*Gurgles*

Re: Correlations and statistics for my webteases

Posted: Tue May 31, 2011 11:03 am
by cactusman
Since people might want to find my original teases, which are rated more highly, I have now made a tag to help find them. The new tag is cactusman-original.

Enjoy!