B Weighing Contribution: Scatter Plot Analysis

[DaveC426913]

I have a list of people whose contribution I'm trying to analyze.
Some people contribute a lot but are off-topic.
Some contribute a little but are on-topic.

How might I weigh them to show an accurate idea of who are the lesser value contributors?

A small sample:

Contributor On~topic Off~topic Total ~ %On ~ %Off
A ~ ~ ~ ~ ~ ~ ~ 79 ~ ~ ~ 49 ~ ~ 128 ~ 61.7 ~ 38.3
B ~ ~ ~ ~ ~ ~ ~ 81 ~ ~ ~ 21 ~ ~ 102 ~ 79.4 ~ 20.6
C ~ ~ ~ ~ ~ ~ ~ 25 ~ ~ ~ 14 ~ ~ ~39 ~ 64.1 ~ 35.9
D ~ ~ ~ ~ ~ ~ ~ ~1 ~ ~ ~ ~0 ~ ~ ~ 1 ~100.0 ~ ~0.0

A contributed 128 comments but 49 of them are off-topic, that's only 61.7%.
D contributed only 1 comment, and it is on-topic, that's a 100% score.

But did D really contribute as effectively as A?

[EDIT]
I guess a scatter plot would work effectively, wouldn't it?

 

[jedishrfu]

Which one am I?

You could do a word count maybe?

 

[DaveC426913]

Heh. Different forum.
The analysis of a thread here would be particularly dull. All posts would be on-topic; all contributors would score 100%.

I did a scatterplot.

I don't think I should post it here - unless I obscure the member names.

But, yeah, a word count would be much better. Some highly on-topic posts are several screens long.

 

[jedishrfu]

Ahh okay then you might be interested in the hero’s challenge on Differential privacy synthetic generation.

They are looking for a scheme to allow researchers to look at real data that has been mangled to make it impossible to identify the customer from the data.

https://www.herox.com/UnlinkableDataChallenge/community

 

[jedishrfu]

You could also rate the reading level of the text and compare scores from the original post ie first is 1st year college followon posts that are of the same level get higher score than ones that too high or too low in grade.

 

[mfb]

Score = Lower bound of Wilson score confidence interval for a Bernoulli parameter?

 

[phyzguy]

How about (# of on-topic contributions) - (some factor) * (# of off-topic contributions). This would reward people for contributing, and penalize the off-topic contributions.

 

[DaveC426913]

Here it is, full-size:
http://www.davesbrain.ca/miscpix/masked-thread-contributors.png