jhnienhuis
07-14-2009, 12:03 PM
Hi,
I've got binary data (0 & 1's) which I represent as a average along a variable "speed". E.g. when speed is 5, the data is [0 0 1 1 0 1 0 1 1 1] and I present that by giving 60% as a input for speed is 5.
Now, the number of binary datapoints varies along the speed, and I would like a confidence interval of any kind to represent accuracy. Treating it as a normal distribution (i.e. do -2 and +2 standard deviation) makes the confidence interval way to high. I thought of adding a fixed number of zero's to do the lower bound and a fixed number of ones to do the upper bound. But what is common for statistics when someone would like to do this?
I've got binary data (0 & 1's) which I represent as a average along a variable "speed". E.g. when speed is 5, the data is [0 0 1 1 0 1 0 1 1 1] and I present that by giving 60% as a input for speed is 5.
Now, the number of binary datapoints varies along the speed, and I would like a confidence interval of any kind to represent accuracy. Treating it as a normal distribution (i.e. do -2 and +2 standard deviation) makes the confidence interval way to high. I thought of adding a fixed number of zero's to do the lower bound and a fixed number of ones to do the upper bound. But what is common for statistics when someone would like to do this?