statwing: a review
TRANSCRIPT
STATWING
• Launched in 2012
• Recent buzz and funding
• Wants to “make your data…dreams come true”
Let’s take it for a spin.
Thomas W. Dinsmore
Statwing offers paid subscriptions at $25 and $100 per month, or a free public license. A paid sub gets you larger files, privacy and the ability to export your charts.
Thomas W. Dinsmore
A two week eval of Statwing’s Silver Plan is free. You do not need to supply a credit card, just an email address.
Thomas W. Dinsmore
Loading files is easy. Statwing supports multiple sources, including Box, Dropbox, Gmail, Google Drive, URLs, Github and your hard drive.
A standard test set of 1,000 rows by 20 columns in an Excel spreadsheet uploaded in a second.
Thomas W. Dinsmore
Oops! One of my standard test data sets (KDD Cup 1998) is too large for a Silver Plan. Statwing knew that in advance (it displays file size as it uploads), but uploaded
85% of it, then prompted me to upgrade. Silly Marketing trick!
Thomas W. Dinsmore
Let’s work with the Bike Sharing Dataset (see notes page for attribution).
Help video narrated by bot.
List of variables here.
Thomas W. Dinsmore
Oops! Statwing can’t read date/time fields. That’s a problem, since this is time series data.
Let’s continue anyway.
Thomas W. Dinsmore
Click! We have a univariate analysis of cnt, a measure that represents the number of bike shares. We’re stuck with that variable name, there is no way to rename it in Statwing.
Thomas W. Dinsmore
Statwing says bicycle shares are “…correlated with season”. WTF? Problem: the seasons are coded 1 through 4, and
Statwing thinks season is a numeric field.
Thomas W. Dinsmore
The solution, per online Help, is to select the Variable * button and change the variable data type. But where is this button? Ha ha, Statwing! Let’s play “hide the button!”
Thomas W. Dinsmore
Statwing tells me that temperature is positively correlated with bike shares. Makes sense. But look at that curve! Can I fit anything other than a straight line? Nope.
Thomas W. Dinsmore
Statwing does two things well: univariate description and bivariate correlation. It can also do crosstabs if it reads your data correctly. (But see slide #10).
Thomas W. Dinsmore
SUMMARY: STATWING
• Nicely executed. Very easy to use. Polished interface.
• Does very little. Narrow use case.
• Lacks elastic pricing.
• Needs more analytic features to justify subscription
Thomas W. Dinsmore
NOTESBike Sharing Dataset courtesy of:
Hadi Fanaee-T
Laboratory of Artificial Intelligence and Decision Support (LIAAD), University of Porto
INESC Porto, Campus da FEUP
Rua Dr. Roberto Frias, 378
4200 - 465 Porto, Portugal
Published in:
Fanaee-T, Hadi, and Gama, Joao, "Event labeling combining ensemble detectors and background knowledge", Progress in Artificial Intelligence (2013): pp. 1-15, Springer Berlin Heidelberg, doi:10.1007/s13748-013-0040-3.
Thomas W. Dinsmore