I added some example analysis code. Actual analysis is detailed below:
Once data is cleaned up (front/back confusions corrected and errors averaged) go ahead and perform an anova and multicompare. The code gives a basic example using fake data.
Make sure to keep track of the number of front/back reversals as well (just add a counter). We will probably use the average number of front/back reversals in a graph to display differences if any between HRTFs. We may wish to perform and ANOVA on this data as well.
Relized I didn't use repeated measures anova. Updated to show new method. Only demos an example at one azimuth. All azimuths should actually be tested.
I added some example analysis code. Actual analysis is detailed below:
Once data is cleaned up (front/back confusions corrected and errors averaged) go ahead and perform an anova and multicompare. The code gives a basic example using fake data.
Make sure to keep track of the number of front/back reversals as well (just add a counter). We will probably use the average number of front/back reversals in a graph to display differences if any between HRTFs. We may wish to perform and ANOVA on this data as well.
Finally, a graph of error frequency and front back confusion frequency is needed: https://epubs.surrey.ac.uk/771637/1/JacksonDesiraju_AES13.pdf in Figure 8