https://www.stat.ncsu.edu/people/davidian/courses/st790/notes.html

I suggest that you look at the notes in the above link. With that many repeats I suggest you look at the models in chapter 6. I hope this helps. Best wishes, David Booth

I agree with David. If the differences between the 16 tasks doe not interest you, for each subject (participant), you could calculate the mean of the 16 recorded times as the dependent variable. As I understand it, you would then have 3 means per subject since they all used the 3 different interaction techniques. In my opinion, you could then run a one-way repeated measures ANOVA to test for the direrences in interaction techniques.

Jos Feys The tasks are important to me. I have already completed a one-way repeated measures anova on the total game completion time for each version. The reason for this analysis is i want to look at the significance between the individual task completion times and the technique used.

Please clarify. You have 16 ‘tasks’. Each task is performed by 3 ‘techniques’ (is this correct??). So there are 48 (=16*3) different task*technique combinations each participant performed. How many participants do you have? Questions that can be investigated are:-Which task is fastest on average?-Which technique is fastest on average?-For task number X, which technique(s) is fastest?-For technique number Y, which task(s) are fastest?

Matthew Robinson Firstly, thanks for your response, it is greatly appreciated. Yes your assumptions are correct. The participants completed the game three times (once per technique). During the game, they were required to complete 16 tasks. Thus for each participant, i have three times for each of the 16 tasks; one for each of the three techniques. I have 20 participants in total. The questions you suggest are some of the questions i'm looking to investigate. The other questions i want to answer are things like whether the order in which they used the different techniques had any impact on performance.

David Booth’s answer is ‘most correct/best’. Since you are probably not a statistician, it might be easier for you to perform and understand a crossover design. Specifically, perform a 3x3-crossover design once per each task (3x3 = 3 period 3 treatments (i.e. 'techniques')). This will let you test for a ‘period effect’ (aka, whether order in which they used the different techniques had any impact on performance.)

With respect to Jos Fey’s response, after you perform the ANOVA, you should run a Tukey’s HSD, which will give you, all the pair-wise comparisons (i.e. tell you which techniques are ‘better’ and by how much)

https://onlinecourses.science.psu.edu/stat509/node/123/

https://onlinecourses.science.psu.edu/stat503/node/24/

-Matt