To Do

From InterSciWiki
Jump to: navigation, search

Early Outline

Nancy Wilkins-Diehr and Ken Hackwork, directors of ECSS at XSEDE want us to continue our Galaxy CoSSci project for a 3rd year, as you know. They'd like to do it as a 3rd year renewal but they can't break the 2-renewal limit. They suggest that someone on the team already be the Principal for a new proposal for this year that would continue our Galaxy CoSSci project for a 3rd year, and that would be you. You kind of own this project already on your CV with the talk at Hamburg, two AA papers, the ABQ paper, and Kepler. 4th year Duran Bell

Being a project director would be great for your CV, whether academic or business. I can write all the bits and pieces of the proposal. You have a user id already at XSEDE. We could do this two ways: either I send those paragraphs to you, you log in at the XSEDE site and initiate an ECSS proposal, same format as our 1st year, upload the paragraphs and the Co-PIs (me and Paul). OR: You loan me your username and password temporarily while I put in the proposal under your name, you get a copy before it goes in, can alter if you want, it is submitted, and then you change your password.

test use of Armin Monecke, Friedrich Leisch 2012. semPLS: Structural Equation Modeling Using Partial Least Squares. Journal of Statistical Software 48#3

What happens in this new year as you know is all the neat stuff like

For HPC on trestles only:

smi <- doMI(dvm, nimp = 60, maxit = 70) for Kepler to increase accuracy
library(bnlearn), expanded by paul's library(bootstrap), 
saving data at XSEDE

NEW FOR CoSSci: DEf01f

mkmappng convex hull maps
Option for Part of Full imputation
Option to increase ski beyond 5:7 and change smi
adding Eff's function to optimize a Lambda λ/(1-λ) exponent for the depvar
expand "Dummy variables" to "Dummy to Normalized variables" where bio.6.n12 rounds bio.6 to a normal curve variate, e.g. 12 easy to map intervals, lowest value 1.0 highest 12.0.
Make all dichotomies 1 2 so they are easy to map.

Comments for students

To see your map(s) click the "eye" button on the third output to Galaxy CoSSci
to Add New Variable use dx$Var to start both Name dx$Var+ and Description dx$Var+<-- ... and put codebook Var in Additional variables to consider
You can make a 1-line complex recode in New Variable like: dx$v2013 <- rnkd(dx$v2013); sub <- rnkd(dx$v2013); sub[which(sub==1)] = NA ; sub[which(sub==6)] = NA ; sub2 <- sub - 1 ; sub2

Don't use Example 2