How accurate is cumtrapz in MatLab?

Say I have a set of discrete acceleration data and want to integrate it to get a set of velocity data. How accurate is the cumtrapz (Cumulative trapezoidal numerical integration) integrator in MATLAB?

Does it follow that the more acceleration data I have (for whatever fixed timescale), the better results I'll get from cumptrapz?

• look up the standard error estimates for the cumulative trapezoidal method (spoiler: for an uniform data grid with spacing $h$, the method is second order accurate, i.e., the error is ${\cal O}(h^2)$). if your acceleration data is non-uniformly spaced, have a look at this scicomp post. Jun 11 '17 at 11:08
• @GoHokies, consider converting your comment to an answer, as it certainly does answer this question completely. Mar 23 '18 at 17:58