I am trying to calculate mEPSC frequency and the smaller amplitude events fall into the noise. Is there a way of calculating a threshold for EPSC amplitude to determine which events should be counted? Thanks.
The article advised by Antoine is worth reading. However, it is a bit " theoretical".
The method implemented there (template search : available on many analysis software) would work better if the signal to noise ration is good. For example if you work with GABAergic events or Glu events of large size (40-100 pA), you will have no problem to extract them. If events are close to the noise (10 pA) it would be more difficult. In my hand I found that the template based search may give an overestimation of mEPSC (false positive) if the noise in high. A trade of is to set a threshold (3* RMS noise) for selection or to inspect each event afterwards which is tedious and somewhat subjective.
Of course that's a problem if you measure the mEPSCs frequency because a reduction in it may result from a decrease in size of the events that would be hidden into the noise. However looking at the size distribution (given that you have a large sample > 500) tell you whether you had a decrease in size (a leftward shift on the histogram) or an actual decrease in frequency.