How to apply a first order filter with a different time constant for each column of a pandas dataframe?

everyone! I have a pandas dataframe where each column represents a noisy signal, and I would like to apply a first order filter (gain = 1, time constant = “x” seconds) in each column, but with a different time constant for each column. For example:

Column_A --> Time constant = 5 secs
Column_B --> Time constant = 15 secs
Column_C --> Time constant = 60 secs
Column_D --> Time constant = 200 secs

Any ideas?

Thanks!

Answer

You can use apply with the args passed as a tuple like so:

def your_filter_func(x, gain, time_constant):
    return some_filter(x, gain, time_constant)

df = df.apply(your_filter_func, args=(gain, time_constant), axis=0)

This applies your_filter_func to each column (that’s the axis=0 part).

If you also want to apply a separate gain and time_constant to each column, you can do this by using the index of the column and adjusting accordingly:

def your_filter_func2(column, gains, time_constants):
    return some_filter(x, gains[column.index], time_constants[column.index])

df = df.apply(your_filter_func2, args=(gains, time_constants), axis=0)

And just make sure that the gains and time_constants are an iterable (e.g. list or numpy array) where you can index into it by the index of the column, to pull out the specific gain and time_constant for that column.