My Photo

April 2014

Sun Mon Tue Wed Thu Fri Sat
    1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30      

BlogStats


« Education, teacher meritocracy, and our national future | Main | edX – ‘a revolution in education’ (addended) »

03 May 2012

TrackBack

TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00e54f86f2ad88330168eb0d74e2970c

Listed below are links to weblogs that reference The Bayes Surface and more (updated 19jan13):

Comments

Douglas Keachie

Seems to me that you could calculate the odds that if you converted every school in the county to charter schools, that the test scores would rise. Of course you'd have fun calculating in the quality of teacher you'd attract with the lower salaries offered, in order to achieve smaller class sizes that supposedly are part of what gives a charter school it's apparent edge. And now that all parents and socio-economic backgrounds would be represented, we'd have a more realistic expectation of overall parent involvement, another factor to be considered.

End result, pardon my intuition, about the same scores as an average, except the poor soci-economic background schools would score even lower, and the upper soci-economic charters would score higher than their public school predecessors. In short, a nice way for the rich and their offspring to get even richer. How nice! Nothing like picking on the weaker members of society to get ahead, survival of the fittest, after all...

Jon

Great stuff!

It would be nice if you'd also show the log likelihood surface, so as to use more of the image on the "interesting" part of the surface when the likelihood ratio is closer to 1. Putting your source up would be nice too (ideally, on github!).

George Rebane

Jon 429pm - thank you Jon, your good suggestion has been incorporated. But I am not sure as to what "source" you are referring.

George Rebane


Jon - I think that I figured out your use of "source" as in source code.  I coded the graphic in Matlab, and the program is rather trivial because it just implements the displayed likelihood ratio version of Bayes rule.  I'm surprised that no one has generated that surface before (at least that I've run across, and I've been messing with Bayes inference for an awfully long time).  gjr

The comments to this entry are closed.

Blog powered by Typepad