Statistics Help @ Talk Stats Forum - Regression Analysis
http://www.talkstats.com/
Linear regression, linear models, nonlinear regressionenTue, 02 Sep 2014 23:08:17 GMTvBulletin60http://www.talkstats.com/images/misc/rss.pngStatistics Help @ Talk Stats Forum - Regression Analysis
http://www.talkstats.com/
Simple Linear Regression with Transformed Variables
http://www.talkstats.com/showthread.php/57471-Simple-Linear-Regression-with-Transformed-Variables?goto=newpost
Tue, 02 Sep 2014 20:36:29 GMTI ran a simple linear regression (length of stay predicts cost). I natural log transformed both the dependent and independent variables.
I am...I ran a simple linear regression (length of stay predicts cost). I natural log transformed both the dependent and independent variables.

I am now attempting to interpret the beta coefficient. I did the following, but have a question about intrepreting it:

e^(beta coefficient*log(1.10). The result is 1.5, so interpretation: given the sample for every 10% increase in length of stay, cost goes up 50%

Question: should this actually read:

For every 10% increase
in mean length of stay, mean cost goes up 50%?

]]>Regression Analysishlsmithhttp://www.talkstats.com/showthread.php/57471-Simple-Linear-Regression-with-Transformed-VariablesSupervised learning on sporting events (Data prep)
http://www.talkstats.com/showthread.php/57451-Supervised-learning-on-sporting-events-(Data-prep)?goto=newpost
Mon, 01 Sep 2014 08:55:07 GMTI'm playing around with the prediction of sporting events, my data looks something like this. A single row for each team and the stats.

Code:

Team StatA StatB StatC
A 4 6 8
B 3 9 8
C 4 6 5

I believe there is two ways it can be setup to build the model, the first would be to translate the data so that both teams who are playing each other are in a single line, so in this example if team A plays teams B then I would just combine those two rows into a single row and there would also be another two columns, HomeScore and AwayScore.

The other way is by matched pairs, so I assign a unique identifier to the teams that are playing each other so my data would look like this

Code:

ID Team StatA StatB StatC
G1 A 4 6 8
G1 B 3 9 8
G2 C 4 6 5

Then the program would know to compare the unique identifiers together. I know SAS can do the matched pair, but not sure if any other program can, rapidminer maybe?

Do these approaches make sense? and should they yield the same result?
]]>Regression Analysisqwirky64http://www.talkstats.com/showthread.php/57451-Supervised-learning-on-sporting-events-(Data-prep)