I have long wondered why the WMF Fundraising department seems stuck on A/B testing instead of multivariate analysis.[1] Does anyone know?
In any case, modern multivariate analysis depends on principal component analysis[2] (PCA) and it really does work great. I have been told for years that it is planned and can't wait to see what happens to fundraising when implemented.
So the reason why I am writing is because I just noticed that EIGENSOFT[3] version 5 has "a new option for PCA projection with large amounts of missing data" in pca.c[4] which is really clean code, too. I hope this helps.
Best regards, James Salsman
[1] http://en.wikipedia.org/wiki/Multivariate_analysis [2] http://en.wikipedia.org/wiki/Principal_component_analysis [3] http://www.hsph.harvard.edu/alkes-price/software/ [4] http://www.hsph.harvard.edu/alkes-price/files/2013/08/EIG5.0.1.tar.gz
Thank you James for this. I totally agree. I recently face PCA and very soon I fell in love with it. In this WikiSym paper we used it to characterise editors of OpenStreetMap based on their temporal activity patterns (eigenbehaivours): http://opensym.org/wsos2013/proceedings/p0207-yasseri.pdf
I'm looking forward to look into EIGENSOFT. cheers, .taha
On Fri, Nov 1, 2013 at 3:40 PM, James Salsman jsalsman@gmail.com wrote:
I have long wondered why the WMF Fundraising department seems stuck on A/B testing instead of multivariate analysis.[1] Does anyone know?
In any case, modern multivariate analysis depends on principal component analysis[2] (PCA) and it really does work great. I have been told for years that it is planned and can't wait to see what happens to fundraising when implemented.
So the reason why I am writing is because I just noticed that EIGENSOFT[3] version 5 has "a new option for PCA projection with large amounts of missing data" in pca.c[4] which is really clean code, too. I hope this helps.
Best regards, James Salsman
[1] http://en.wikipedia.org/wiki/Multivariate_analysis [2] http://en.wikipedia.org/wiki/Principal_component_analysis [3] http://www.hsph.harvard.edu/alkes-price/software/ [4] http://www.hsph.harvard.edu/alkes-price/files/2013/08/EIG5.0.1.tar.gz
Wiki-research-l mailing list Wiki-research-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wiki-research-l
wiki-research-l@lists.wikimedia.org