| Reference | Downloads | Github

Response time accuracy

Hi, we have an assignment to make a classic stroop expirement, and ask to present the "error bars"
I not sure, but i think is the accouracy of the program to calculate the exact time of the response
so, can you tell me what is the accouracy of the time?

and if you think the error bars should be somthing else- let me know :wink:

thanks a lot! Viki

Most likely, they just want you to present the RT +/- confidence interval or Standard Error (of the mean). Response times are very accurate to below 0.1 ms on the software-side in psychopy. But on many keyboards there’s a buffer which causes up to 30 ms in jitter. This would apply to all stimulus software.

Note though, that reaction times are not normally distributed, so confidence intervals or standard errors are really poor models since they assume normal distribution. Draw a histogram and see for yourself. And go show your teacher - I’m betting that he/she haven’t thought about it :slight_smile: Sometimes RTs can be close to log-normal, i.e. log(RT).

So medians and interquartile-range is usually the best way to represent reaction times.

If you check for keyboard responses at the end of each frame you’re limited to the monitor’s resolution (at 60Hz that would be 16.66ms).

Independently of that you should also display a range of “uncertainty” for your reaction time estimates. @lindeloev’s suggestion with median+IQD is very good, if you also want to include outliers you should create a whole Box plot.

The post is requesting how to present error bars (std err). This is a matter of statistics based on the actual responses of participants. I’m sure the student’s tutor does not mean present an error bar with a “theoretical” error range on.

This isn’t a relevant forum to be discussing how to add error bars.