How to set values greater than Threshold to NaN

Might be faster, I don't know. At least the Extract operation runs internal C++ code, and using the index wave means iterating over fewer wave elements. If the waves are large and the over-threshold values sparse, this could be the way to go.

I'd use John's first response, possibly with the Threadsafe keyword if you really need the performance. Either way I don't think this is going to be a bottleneck in your analysis.

I will caution you about introducing NaN values into data. Make sure that you pay attention to how the various Igor commands you use handle NaNs. Some of them will ignore NaNs and some will not allow them. For some operations (e.g. Smooth), NaNs in the data might give undefined results. Also make sure that you use numtype() if you want to test a value for NaN. You can't just do something like if (x == NaN)... since that sort of comparison will always be false.

Assuming that the previous post did not convince you out of using NaNs, I'd recommend using MatrixOP.

MatrixOP/O modifiedWave=setNaNs(inputWave,greater(inputWave,threshold))

The advantage here is that it would create the appropriate output wave (something that can hold a NaN) even if your inputWave is of integer type.  It is also faster than executing the command with the '?' operator.

AG

Note that MatrixOp is very efficient, but does not transfer any scales to the output wave. You might want to use 'CopyScales inputWave, modifiedWave' afterwards to transfer the scales.