Screening of Variables
The “Screening of Variables with T-Test” assesses whether the averages of two groups are statistically different or not, and extract variables which distinguish target groups from the others. It calculates “differential score” from mean ratio and t-test p value, and shows them with box-whisker plots and ROC curves.
Go back to sheet “Appliances” and click “Multibase_2015” and “Open Form”. Then after Preparation dialog box opens, check “Screening for Variables with T-Test” and click “Next >>”. Then new sheet “Multibase_Preparation” will be generated.
Click “Multibase_2015” and “Open Form” again to open "Screening Variable" dialog box. In this box, you need to define target groups and control groups, which you want to compare. Just highlight California and Florida and click “Set as Cases(T)”, and select Massachusetts, Pennsylvania and Washington and click “Set as Controls(C)”. If Multibase recognizes them correctly, their labels are changed with "T" and "C" as shown below.
Next, click “Done”. Variables which distinguish target groups from control groups are extracted automatically and shown with box-whisker plots, ROC curves.
A box-and-whisker plot depicts distributions with five summaries; the minimum value, lower quartile, median, upper quartile, and maximum value. A receiver operating characteristic (ROC) curve shows graph of the sensitivity vs false positive rate (1 − specificity) for evaluating variable performances.
Variable map shows how variable contribute differentiation between two groups. X axis indicates average ratio between two groups and Y axis means –Log(P). The upper-right and left-down corner is most significant, meanwhile the variables around origin do not contribute differentiation..
2-3-1 Marunouchi Chiyoda Tokyo Japan