Dear All,

I want to do a feature selection for predicting the kinematics characteristics (we have a total of 9) of tennis shots that better predict an outcome variable called quality of the shot (with three values, 1, 2 or 3, I think it is better not to go into details). I think it is a good idea to use ANOVA to select features (below I add some references that use this test to select variables for creating a prediction model) . But I have a question… My independent variables (kinematic variable 1, 2,..., 9) are continuous and my dependent variable (quality of the shot) is ordinal. It is correct if we select the better predictive variables comparing the kinematics variables for the three quality level of shots? (ANOVA comparing kinematic variable 1, 2, 3,...,9 between shot of quality 1, shot of quality 2 and shot of quality 3). Or it is better to categorize the independent variables (for example speed of the hand could be categorized as slow, medium and fast using terciles) and then compare the dependent variable (level of the shot) in those level of the independent variable (slow, medium and fast hand movement) using for example Kruskal Wallis.

Other ficticious example to make my question more undenstable. Imagine that we want to predict which of three anthropometric characteristics predict the level of the basketball player (beginner, competitive or elite): height, hip height or wingspan. It is correct to compare this characteristics between those three groups and select the better predictive variable of the level selecting the variables with the higher F values? Or it is better to categorize the independent variables (height, hip heigh and wingspan) and compare in each one the dependent variable with kruskall Wallis and then selecting the minor p values?

Thanks very much,

Gabriel

References:

Yamashita, Alexandre Yukio, et al. "The Residual Center of Mass: An Image Descriptor for the Diagnosis of Alzheimer Disease." Neuroinformatics (2018): 1-15.

Chen, Yi-Wei, and Chih-Jen Lin. "Combining SVMs with various feature selection strategies." Feature extraction. Springer, Berlin, Heidelberg, 2006. 315-324

Surendiran, B., and A. Vadivel. "Feature selection using stepwise ANOVA discriminant analysis for mammogram mass classification." International J. of Recent Trends in Engineering and Technology 3.2 (2010): 55-57.

Elssied, Nadir Omer Fadl, Othman Ibrahim, and Ahmed Hamza Osman. "Research Article A Novel Feature Selection Based on One-Way ANOVA F-Test for E-Mail Spam Classification." Research Journal of Applied Sciences, Engineering and Technology 7.3 (2014): 625-638.

Zaki, Wan Mimi Diyana Wan, et al. "Automated pterygium detection method of anterior segment photographed images." Computer methods and programs in biomedicine 154 (2018): 71-78.

Ali Khan, Sajid, et al. "Kruskal-Wallis-based computationally efficient feature selection for face recognition." The Scientific World Journal 2014 (2014).

More Gabriel Delgado García's questions See All
Similar questions and discussions