Friday, February 24
PS2 Poster Session 2 and Refreshments Fri, Feb 24, 5:15 PM - 6:30 PM
Conference Center AB

Analyzing Shot Data with MANOVA (303454)

*Victoria Cox, Dstl 

Keywords: MANOVA, shot data, R

Shot data, or X and Y coordinate data in general, is commonly treated as independent for ease of analysis; however this is a large, and mostly incorrect, assumption to make. In addition most methods used to analyze this type of data deal with independent standard deviations and averages. Other commonly quoted outputs include extreme spread plots and Circular Error Probable (CEP) plots which aren’t necessarily the best for providing useful information.

Through this poster I will demonstrate the use of Multivariate Analysis of Variance (MANOVA) to deal with the correlation between the X and Y response variables whilst still testing for significant differences between explanatory variables. I will also highlight the assumptions of MANOVA, describe the limitations of using the methods and plots listed above, and show the benefits of confidence ellipses to display the data graphically. The R output will be translated for example communication to a customer to provide evidence for decision making. Finally, I will pose some issues with sample size calculations for MANOVA data for which I seek advice.

The example used will be a real scenario with dummy data; it will contain X and Y shot vector data for a number of different weapons and ammo types. The aim is to investigate significant differences and ascertain the accuracy and precision. The analysis will have been conducted in R v3.2.2 and all packages used will be listed.