Interesting stuff. Among many other claims are that it will find “Incorrectly identified principle components”. Still need a spelling checker, then.
(Yes indeed, that really was pedantic.)
More to the point, Lee Wilkinson has such a track record that it can’t be snake oil, despite the sales pitch.
In stats courses for non-statisticians, I often argue that most such courses and online material seemed designed to enable people to do their own statistics without giving away that they have no idea what they are doing (and especially why).
AdviseStat seems to take that to an extreme and I am starting to think that is mostly a good think (typo purposely left in – thing?). Being like automatic translation which at some point (my wife no longer works in that field) will do more good than harm for the revisers who finalize and approve them (they used to just throw away novice and automatic translations and start over.)
Unfortunately all the pre-testing specification searches, multiple choices and Ed George’s “whose posterior” should it be stuff – will undoubtedly get in the way.
Neat commercial!
“would AdviceStat have saved Romney from ….. No it would not ” kkkkkk
I think they torpedo their own software by agreeing to the statistic, (merely pointing out some possible causal explanations for it) and amplifying the all-too-trite/immature obnoxiousness of the day.
I have a license for AdviseStat, partly because I think Wilkinson’s stuff is usually worth looking at (I bought a copy of SYSTAT from him back in the CPM, 5.25 inch floppy days and have been a beta tester) and partly because I was looking for low level, simple text analysis that I could share with classes without getting in over their heads. (mine, too — that’s not my area).
Each analysis comes with header information providing at least some theory and advise about what the heck things mean — download the white papers at the site to see some examples of it.
I don’t know whether the market for this type of thing is big enough to make a commercial go of it, but at least he’s not just writing the 5000th introductory book on how to use R.
Interfaces, interfaces, interfaces… are they converging?
It seems they are getting closer to be fool-proof.
Using simple tools like this, I imagine that analysis will
become more democratic and only more sofisticated analysis
will remain for the statisticians. This can lead to a
reduction in the demand for statisticians. On the other hand,
once everybody become proficient with this tool, market forces
will ask for more, what will demand more skilled statisticians
and, therefore, more rigorous training.
And, of course, it is good to see this tool comming from
the statistics community instead of the computing community,
as usual.
Interesting stuff. Among many other claims are that it will find “Incorrectly identified principle components”. Still need a spelling checker, then.
(Yes indeed, that really was pedantic.)
More to the point, Lee Wilkinson has such a track record that it can’t be snake oil, despite the sales pitch.
In stats courses for non-statisticians, I often argue that most such courses and online material seemed designed to enable people to do their own statistics without giving away that they have no idea what they are doing (and especially why).
AdviseStat seems to take that to an extreme and I am starting to think that is mostly a good think (typo purposely left in – thing?). Being like automatic translation which at some point (my wife no longer works in that field) will do more good than harm for the revisers who finalize and approve them (they used to just throw away novice and automatic translations and start over.)
Unfortunately all the pre-testing specification searches, multiple choices and Ed George’s “whose posterior” should it be stuff – will undoubtedly get in the way.
Neat commercial!
“would AdviceStat have saved Romney from ….. No it would not ” kkkkkk
I think they torpedo their own software by agreeing to the statistic, (merely pointing out some possible causal explanations for it) and amplifying the all-too-trite/immature obnoxiousness of the day.
I have a license for AdviseStat, partly because I think Wilkinson’s stuff is usually worth looking at (I bought a copy of SYSTAT from him back in the CPM, 5.25 inch floppy days and have been a beta tester) and partly because I was looking for low level, simple text analysis that I could share with classes without getting in over their heads. (mine, too — that’s not my area).
Each analysis comes with header information providing at least some theory and advise about what the heck things mean — download the white papers at the site to see some examples of it.
I don’t know whether the market for this type of thing is big enough to make a commercial go of it, but at least he’s not just writing the 5000th introductory book on how to use R.
Interfaces, interfaces, interfaces… are they converging?
It seems they are getting closer to be fool-proof.
Using simple tools like this, I imagine that analysis will
become more democratic and only more sofisticated analysis
will remain for the statisticians. This can lead to a
reduction in the demand for statisticians. On the other hand,
once everybody become proficient with this tool, market forces
will ask for more, what will demand more skilled statisticians
and, therefore, more rigorous training.
And, of course, it is good to see this tool comming from
the statistics community instead of the computing community,
as usual.
Pingback: Pronostics électoraux sur l’élection américaine | Polit’bistro : des politiques, du café