TheSharklord wrote:CyLaw wrote:TheSharklord wrote:I think it draws all of its data from LSN. It's the same as looking at the numbers yourself, with the provision that to do it as well as LSP you'd have to be able to look at all the data simultaneously, process decision trends, generate theoretical LSAT/GPA weights, etc. Which is hard to do with any unbiased accuracy.
Anyhow, like any statistical model, it's only as good as your data. And for URMs and heavy splitters, and goodness knows the intersection of the two (which I am), there's not as much data to support the conclusions because you're on the edge of the numbers.
Anyhow, while LSN may not be perfect for individual data points (w/ flames, fudged numbers, etc.) on the whole, there are a ton of data by now so the trends are fairly strong if you're a "standard" applicant...
My 2 cents.
It does not draw its data from LSN. The LSN data was only used for accuracy testing and refinement, AFTER the models were created.
http://www.lawschoolpredictor.com/?page_id=173From LSP Site wrote:Law School Predictor is of far greater use to applicants when applicants can gauge the accuracy of LSP predictions. These results are based off of more than 33,500 Law School Numbers-listed law school admission decisions from the 2008-09 cycle for all ABA schools listed on both LSN and LSP. The version of LSP tested was Version 2.4. A big thanks goes out to TLS forum member CyLaw for compiling the data from LSN; I couldn’t have done it without CyLaw.
Not sure I follow. That pretty much reads as the LSN being the data source to me. Insofar as data and statistical modeling are inextricably tied together (with no data a model is as good as a theory with no proof, made without any natural observation--ie nothing at all), I think I disagree, unless I have a fundamental misunderstanding of how LSP works (which is always possible--say, "Weak Consider" =P)
It might help if you look at the title of the page you quoted that from, namely the page Accuracy, which was the results of an accuracy analysis ran a year after the creation of LSP. I know this for a fact, as I was the one who supplied the LSN data a year AFTER the first LSP was initially created, so the models were not created from LSN data, as LSP did not have such data until I gave it to them after they were created.
Edit: You will have to ask YC where the initial data came from for LSP, but it was not LSN.