SMLG (Statistical Machine Learning Group) Discussion Forum

bakerchowdhury wrote:Hello everyone:
Attached is my bivariate analysis report for the project. I will be selecting variables using bivariate results for model building.
Baker

As stated in the instruction of the Final Project Report Wiki page, you should update your Final Project Report Wiki page in Blackboard to include weekly outcome that you will post here. Please read carefully the instructions and follow them carefully. What is the number of the total sample that you will be using for this project (it is reported that almost 81% of answers needed to assess the outcome is missing)? Also you should be putting more efforts into the class project. You should be reporting NBC and/or logistic regression results by now.

oosib002@fiu.edu wrote:Please see attached the outputs for logistic regression and NBC for the outcomes (Cigarette use & Bullying, Tobacco Use & Bullying and Drug use & Bullying).
The outcomes were paired with bullying at school and electronic bullying separately for the 3 models included in the analysis.

Thank you,
Ola

Great start. Please work on implementing calculating IC, log (natural) likelihood, and training error for NBC and logistic regression and report them with the model.

cdaws011 wrote:Hi All,

Please see attachments for my associations table (chi-square analysis), NBC output, and logistic regression output. All analyses were done using R.

Great start. As I replied to Olatokunbo's posting, please work on implementing calculating IC, log (natural) likelihood, and training error for NBC and logistic regression and report them with the model.

Attached is my results from forward selection method

Please see attached the loglikelihoods, AIC/BIC and training errors for Log R and NBC.

Thank you,
Ola

Attached are the outputs for forward, backward and step wise selection methods

Hi Everyone,

Attached is the NBC and LogR model comparison with log likelihoods and AIC along with my R code. Sorry for the late post. I was having trouble running the R code. I kept getting the following error messages while I was trying to fix the problem: invalid to change the storage mode of a factor, Error in eval(expr, envir, enclos) : y values must be 0 <= y <= 1. I ended up converting my SPSS file into a csv file and recoding the variables using reference coding. I also had to remove extra cases that had remained with missing data when I had split the dataset during data cleaning. After that, my R code ran fine without any problems. I'm still trying to do the ROC curves and the cross validation in R. I hope to post those very soon.

Best,
Christyl

Attached is my class project presentation.

Below attached are all the file pretaining to the project on comparison of NBC with LogR using GBM data.

bakerchowdhury wrote:Attached is my class project presentation.

Please present descriptive statistics about your input and outcome variables for your final model. Also describe how you categorized any continuous variable for NBC.