Calculating stderr for a bar plot of rate of events detected I'm trying to understand the correct way to calculate standard error bars for the plot below.
rate:    10 |
events      |     _______
per       5 |     |     |
minute      |_____|     |______
          1 |     |     |     |_____
            |_____|_____|_____|_____|
           0.0   0.25  0.5   0.75  1.0
            measure of detected motion

On X we have a measure of motion in a video, looking at each second of video over time, each 1s falls into a range of [0, 1] of measured motion. We are counting how many events of a particular type occur relative to the computed motion at that time.
So for Y we have a count of events per minute. It's normalized by time spent observing each bin of motion.
We have 15 total subjects being observed over a long period of time.
Questions:

I'm having trouble figuring out how error bars should be represented
on data like this. It's not quite a histogram because we normalized
event count by time.
Also, I'm not sure if I should be using 15 subjects as the denominator
in stderr calculations, or total events, or total time in a bin.

An example of data would look like:
second, bin, event-detected
  1      0        no
  2      0        no
  3      0       yes
...
 100     2        no
 101     2       yes
 102     2        no

 A: So for each 1 second interval the endpoint is a number between 0 and 1?  Is the set of possible values discrete, e.g. 0, 0.1, 0.2, ..., 1, or continuous?  Based on your brackets it looks like 0 and 1 are inclusive.  It appears that you are constructing a histogram and if the endpoint is continuous it can be conceptualized as random draws from a beta distribution.   You wouldn't put standard error bars on a histogram, but you could on an empirical cumulative distribution function (ECDF).  Alternatively you could fit a beta model to the data and plot the estimated beta CDF with standard error bars or with tolerance intervals (confidence intervals for population percentiles).
If the endpoint, $Y$, is measured in discrete units, such as multiples of 0.1, you could conceptualize the endpoint $10Y$ as random draws from a binomial distribution.  The options of using a histogram, ECDF, and estimated parametric binomial CDF with standard error bars or tolerance intervals would apply here.
If the raw data involves an endpoint that is bounded on the left by 0 and essentially unbounded on the right this could be conceptualized as random draws from a Poisson or negative binomial distribution (discrete) or a gamma distribution (continuous).
