Logistic Regression for Funding Factors

We ran logistic regression on about 100 different variables ranging from the prefix of the projects teachers name e.g. Mr. or Ms. to the cost of the project. The purpose was to attempt to predict the success of the project i.e. whether it achieved full funding based on the known variables when the project was purposed. Logistic regression learns weights for each different variable in order to makes its prediction, essentially determining which variables are important factors for predicting the future success or failure of the project. The importance of the variable is expressed as an odds ratio, that is, the probability of success given the information the variable provides divided by the probability of failure given the same variable. Basically, an odds ratio of greater than one indicates the variable is generally indicative of a successful project whereas a value of less that one means the opposite. The odds ratio for all the variables along with the raw output from Weka (the software used to run logistic regression) is attached to this page, however we are presenting the 25 most significant variables. The results from the logistic regression analysis generally agree with the results of the clustering analysis, that is, the Southern states appear to be at significant disadvantage to the Northeastern and Western states. Also, projects requesting books and trips are stronger than those requesting technology As for subjects, Sports, Economics, and Nutrition outweigh Character Education, Early Development, and Parent Involvement. Finally, projects with matching donations were some of the strongest. Finally, logistic regression was able to accurately predict the success or failure of a project 72% of the time.

Top 25 Most Significant Variables

In short, the values greater than 1 mean it indicates success whereas values less that 1 mean the opposite.