Best regards.

First, I would definitely not use a bar plot for the time series. With a line it is much easier to see the connection between days than with a bar plot. A stem/leaf plot might work, though, so I would try that out. It is like a bar plot, but instead of wide bars it uses thin lines with a circle (or other shape) at the peak. It is fairly easy for the eye to connect the dots into a curve but it does not look quite as jaggy as the line does. I would also experiment with different time periods (6 hours, 12, hours, 1 day, 2 days, etc). Perhaps post different combinations next month (or this month even) and let there be some discussion about what seems useful and what doesn’t. You can then narrow it down for future issues.

First, I think doing a 3 and/or 5-day sliding moving average would help smooth at some of those bumps (so that each day is actually the average of that day and the 1 or 2 days before or after). This would not replace the time series, only complement it.

Second, I would have a histogram for the number of ideas for each user. So for instance the number of users with 1 idea, the number of users with 2 ideas, the number of users with 3 ideas, etc. That will tell us whether are dealing with a few users posting a lot of ideas or a lot of users posting a few ideas.

I would also post the number of ideas with each tag, at least for the top 10-20 most popular tags.

Posting a time series for comments on ideas would also be useful, to see how active discussion is and how it is changing over time. Perhaps overlay it on the same plot as the time series for ideas, making comparison between the two. You can have plots with two y-axes, one on the left and one on the right. This would allow you to scale the two time series plots so they are similar in height even though their true range is very different.

Another interesting thing would be a plot with the number of comments on one axis and the number of positive votes on the other (a vote vs. comment count plot). This would allow us to see whether commenting and voting are related, and if so in what way. Perhaps with a correlation coefficient. Same with comments vs. controversy.

Having a histogram for the number of comments per idea (number of ideas with 0-5 comments, number of ideas with 5-10 comments, etc.) would be interesting. So would a histogram for the votes and a histogram for the controversy. Since there are far more votes and comments than there are ideas those would probably have to be broken into bins (number of ideas with 0-5 votes, 5-10 votes, etc).

A pie chart representation of the ideas in each category, pending, implemented, done, etc., would be a good visual representation of that data.

The count of ideas for each of the top 5-20 most prolific forum members, the count of votes for the top 5-20 highest-rated ideas, and the controversy level for the top 5-20 most controversial ideas would also be interesting, along with links to those ideas. You could also post the text of those 15-60 ideas at the end, like how the commit digest posts the labels for commits at the end.

Those are all the potentially interesting ideas I can come up with right now. I will post more if they come to me.

Regarding the graph, I like the line graph – it shows a trend, the relationship between time and number of ideas. I don’t know which type is the “right” one, but that’s what I personally think.

There is one proposal for presenting number of votes per day – because they are discrete numbers spread over discrete timeline, you should display votes through histogram with constant bar width with no space beetwen them. One such chart can be found on 2nd page of document at:

http://www.cdc.gov/healthyyouth/evaluation/pdf/brief12.pdf

Keep on rocking with KDE ;)

Thumbs up!

@aki For now I’d like to see the reactions first. I’m still not sure it’s Dot-worthy material.

@all: thanks a lot!