InfoDump data manipulation? June 15, 2011 5:31 PM   Subscribe

Is there anyway to get a chart of my comments over time, using the Infodump and Excel (or any other tool suitable for the job).

Sorry if this is more AskMetafilter related, but I thought it was Metafilter related enough to go in here, as other Metafilteranians might want to try this.

I'm looking for something comparable to what you see with website statistics, like the traffic chart here. I went to the wonderful wiki guide, as I haven't touched Excel in a very long time. I have the info dump filtered to my user id, simple enough.

How would I go about creating this chart? Would I need a way to match up my comments to all days since I've been here? Otherwise periods of not using the site wouldn't show up, right? I'm assuming that this is a common enough scenerio that something like Excel would have a built-in way of doing it. I'm also a bit surprised, based on previous infodump threads, that this has not come up before.
posted by geoff. to MetaFilter-Related at 5:31 PM (10 comments total) 1 user marked this as a favorite

So you want the number of comments per day, or some other time measure? Week? Month?

I can't help you do anything in Excel, but I can write something in Python and get you a simple count of comments per day, if that's what you want. I'm a little sluggish right now, though, so someone might have a better solution before I get this done.
posted by -jf- at 6:05 PM on June 15, 2011

I've wanted this for the longest time ever since I discovered TweetStats.

This would be such a pretty pony.
posted by desjardins at 6:34 PM on June 15, 2011 [1 favorite]

Ok, so, I have CSV files (tab-delimited) of comments per day and comments per month. If that's helpful to you, geoff., let me know and I'll link to them.
posted by -jf- at 7:11 PM on June 15, 2011

I think that using PivotTables and PivotCharts would make this trivial. Highlight your data and play around with those; I think you'll be able to create what you're looking for.
posted by emumimic at 8:52 PM on June 15, 2011

One easy way to get the per-interval stats is to convert the date/time field to unix epoch time (integer seconds since 1970-01-01) and then calculate 'floor(value / interval) * interval' for each value, which 'bins' the values to the interval, which you then just plug into a histogram function. Or you could substring functions on ISO date formats to similar effect I suppose.
posted by Rhomboid at 9:24 PM on June 15, 2011

as other Metafilteranians might want to try this

How can anyone think about charting their comments when we've got an amazing word like "Metafilteranians" sitting there on the screen unremarked upon? We should all be collaborating on reworking this into "Metafilteranian Homesick Blues."
posted by amyms at 10:57 PM on June 15, 2011

I prefer "Metards".
posted by the quidnunc kid at 1:47 AM on June 16, 2011

( ... Which appears to have been coined by 23skidoo)
posted by the quidnunc kid at 1:51 AM on June 16, 2011

The site don't work
Cuz the hool'guns stole the JRuns

posted by cortex (staff) at 6:57 AM on June 16, 2011

The man in the bicycle cap
By the big door
Wants five dollar bills
You only got four

posted by amyms at 9:38 PM on June 16, 2011

« Older May the road rise to meet you.   |   Double Double Toil and Trouble Newer »

You are not logged in, either login or create an account to post comments