Analysis of Human Reaction Times

[danv_MIT]

Hello,

I'm conducting a science fair project on human reaction times- I want to determine how each of several factors affects reaction time. I have over 50 subjects who took my reaction time test (a computerized assessment, accurate to the millisecond). Each subject was measured 50 times on their latency to respond to a simple stimulus (box appears on-screen).

I recorded each subject's age and gender. In addition, I had them answer some questions to determine a numerical value representing their physical fitness, mental ability, fatigue, and stress levels. I also have the individual reaction times for each subject-- all 50 of them, not just the averages. (I knew I might need to compute standard deviation or something similar so I didn't discard the individual times.)

Now, after having collected the data, it seems I'm not so sure about how to draw conclusions from it. I want to know how each of the aforementioned factors affects the reaction times-- which affects it the most? Which has little effect on the reaction times?

I'm enrolled in an AP Statistics class, but we haven't studied ANOVAs in depth yet and I'm not sure whether ANOVA is the best choice for analyzing this data/answering these questions. Are there other superior methods of analyzing this kind of data? Any help would be appreciated.

(I'm not sure whether this post belongs here or in the math/programming section... )

Dan V.
MIT Hopeful

[donnahardy2]

Hi Dan,

This sounds like an excellent project. Your question would probably best be answered in the math/programming section so please do post your question there. However, I believe that one way ANOVA would be a suitable method and would allow you to compare results by gender, age, physical fitness, etc. Since your data collection included 50 measurements for each individual, you will be able to calculate the standard deviation needed for this analysis.

http://en.wikipedia.org/wiki/Analysis_of_variance

http://www.physics.csbsju.edu/stats/anova.html

Good luck!

Donna Hardy

[att159]

I've done behavioral projects for the last couple years and I've used z-tests both times - not sure if that would apply as well here. You could check out a book on hypothesis testing.

You could also ask your AP Statistics teacher, or a science teacher.

[MelissaB]

Hi,

I also read your post over in the math section. You should use a general linear model, which will allow you to include fixed effects (gender) and covariates (age) on the averages of each of the 50 people. What statistics program are you using?

In terms of correlation and causation--yes, this is an entirely correlative study. You did not experimentally manipulate age or gender or... However, with human behavioral studies (and with behavioral studies in general), this is the norm. There are many factors out there that we cannot control and we cannot experimentally manipulate--I personally work on the effects of weather on birds, and weather definitely cannot be manipulated in the wild!

So, it is good that you realize that this is not a 'perfect' test and that something else that's correlated with age, gender, stress levels, etc. might actually affect reaction time--but this is OK. It's just something to keep in mind as you speak to the judges.

In terms of trying to know exactly which variable affects reaction time 'most', this is a little trickier. I would suggest doing partial regressions--but this is not ideal if you have variables that are highly correlated with one another. Have you looked at the correlations between age, stress levels, etc.? Post back with those and then I can advise you more.