Software R

From InterSciWiki
Jump to: navigation, search

In our fall 2010 class we will use open source R programs to get causality estimates for the SCCS or other data. They are loaded in the 155 SST lab and can be installed on PCs or Macs for home use - see Imputing_data_for_Regression_Analysis#fall_2010, includes netbook linux. Work starts on the wiki at EduMod-71 with three programs that can be run separately or all at once, and changed to adapt to your topics of interest.

SCCS R package Install - already done at our lab, can do at home e.g., EduMod74:
source("examples/src/create_model_value_DRWchildren.R") #create_model_value_children.R -- here you
*edit the dependent variable and delete it from the independent variables for 'run_model.R' below.
source("examples/src/run_model.R") #takes whatever model is currently defined, does multiple imputation, 
*calls two stage ols and saves results to file on the working directory (.../sccs).

You can copy the code from EduMod74 (or the older versions of EduMod-71+pages]]: --> and put your source files into "Working" or "Notworking." Use your own EduMod- for the User:page that comes with login, test an example, copy the results to your page, and verify the programs are working. Then you choose a dependent variable from the codebook, change that one element in the program, and the results will show how only certain of the SCCS variables affect your dependent variable. There are 1060 published articles on the SCCS in Google Scholar. You will be able to find articles by adding "+ Your_topic_word" to the scholar search. In your paper you can compare what is known about Your_topic and what you are discovering in class. We compare notes and findings with powerpoint presentations starting in week 6. Each of these steps help to form your writing project.

Edu-Mod 2009-10: The Individual Studies -from two previous classes

Your findings are guaranteed to be new and original because they are much more accurate than estimates of the effects of variables on one another that do not take societal interactions and common histories into account. If the R program shows errors, check what to do to Debug R, ask or email your instructor, or for last year's software check Error messages for Eff and Dow 2009. (Probably best not to use the older EduMod software.

If you want to do a comparison of your results with results using SPSS, available in our lab, using multiple regression WITHOUT considering interactions, common histories, or imputation of missing data, I will be illustrating the findings and weaknesses of that approach for you some weeks from now in class.

These approaches and the software will transfer to all kinds of new career paths in the new economy. You can take the open source code with you - it is a free resource.

Links

SCCS R package SCCS R package Install