A repository for collaborating on our new manuscript investigating how keystroke dynamics conform to information theoretic measures of entropy in the letters people type.
Once we figure out our analyses for determining how much variation in letter uncertainty explains variation in keystroke times (for position and word length), we can move onto the repetition problem.
In choice reaction time experiments (like Hick-Hyman law), it was noted early on that changing the probability of specific choices also results in changing the proability of getting repeated sequences in task. For example, if you have four options, and option 1 and 2 are happening alot, then trial n-1 to trial n repeats (11 and 22) are also happening more than they otherwise would if the four options were random.
It is well known that reaction time can be faster on repetition trials compared to alternation trials, so there was some suggestion that RT was not tracking information per say, but instead the relationship between RT and H might simply be a result of averaging over conditions with different proportion of repeated trials.
We should be wary of the possibility that the same repetition issue could be at play in our analysis. I can think of some ways to address the issue, but for now I will leave the issue open.
How can we address the issue? Can our evidence that letter uncertainty over position and word length explains variance in keystroke typing times be "explained" by repetition priming? How could we assess this possibility?
Once we figure out our analyses for determining how much variation in letter uncertainty explains variation in keystroke times (for position and word length), we can move onto the repetition problem.
In choice reaction time experiments (like Hick-Hyman law), it was noted early on that changing the probability of specific choices also results in changing the proability of getting repeated sequences in task. For example, if you have four options, and option 1 and 2 are happening alot, then trial n-1 to trial n repeats (11 and 22) are also happening more than they otherwise would if the four options were random.
It is well known that reaction time can be faster on repetition trials compared to alternation trials, so there was some suggestion that RT was not tracking information per say, but instead the relationship between RT and H might simply be a result of averaging over conditions with different proportion of repeated trials.
We should be wary of the possibility that the same repetition issue could be at play in our analysis. I can think of some ways to address the issue, but for now I will leave the issue open.
How can we address the issue? Can our evidence that letter uncertainty over position and word length explains variance in keystroke typing times be "explained" by repetition priming? How could we assess this possibility?