[Bug] Incorrect Output in `time_histogram` for Single Unit Spike Train with `output='rate'`

[utwrd]

Description
When using time_histogram with output='rate', the result differs between using a single spike train (time_histogram(spiketrain, output='rate')) and wrapping it in a list (time_histogram([spiketrain], output='rate')). The correct result is obtained with the latter (time_histogram([spiketrain], output='rate')), but this discrepancy is confusing when we want to calculate the firing rate for each unit.

Expected Behavior
Both calls to time_histogram should yield the same result, regardless of whether a single spiketrain or a list of one spiketrain is provided.

Actual Behavior
time_histogram(spiketrain, output='rate') produces an incorrect result, whereas time_histogram([spiketrain], output='rate') provides the correct output. This discrepancy is due to how the function calculates the sample length based on the input type.

Cause
This issue arises from a condition in statistics.py, line 1191, where len(spiketrains) is used. When spiketrains is not provided as a list, len(spiketrains) mistakenly returns the length of the spike train sample rather than counting the number of spike trains, leading to an incorrect rate calculation.

Suggested Solution
To fix this issue, modify the code to ensure that spiketrains is always treated as a list, even if it contains only one unit. This can be done by wrapping spiketrain in a list if it is not already, ensuring consistent behavior for both single and multiple spike train inputs.

Steps to Reproduce

1. Create a single spike train, spiketrain.
2. Run time_histogram(spiketrain, output='rate') and observe the incorrect result.
3. Wrap spiketrain in a list and run time_histogram([spiketrain], output='rate').
4. Compare the results; the correct output is produced only when spiketrain is wrapped in a list.

[Moritz-Alexander-Kern]

Hey @utwrd ,
thanks for reporting this, I was able to reproduce this with the following minimal example:

import quantities as pq
from elephant.statistics import time_histogram
from neo.core import SpikeTrain

# Create a single spike train
spiketrain = SpikeTrain([0.1, 0.5, 1.0, 1.5, 2.0] * pq.s, t_stop=3.0 * pq.s)

# Run time_histogram with spiketrain directly and observe the incorrect result
histogram_direct = time_histogram(spiketrain, output='rate', bin_size=0.5 * pq.s)
print("Histogram (direct):", histogram_direct)

# Wrap spiketrain in a list and run time_histogram
histogram_wrapped = time_histogram([spiketrain], output='rate', bin_size=0.5 * pq.s)
print("Histogram (wrapped):", histogram_wrapped)

Which yields:

Histogram (direct): [[0.4]
 [0.4]
 [0.4]
 [0.4]
 [0.4]
 [0. ]] 1/s
Histogram (wrapped): [[2.]
 [2.]
 [2.]
 [2.]
 [2.]
 [0.]] 1/s

Using elephant == 1.1.1.

This example confirms that the result for a single SpikeTrain is incorrect when using the time_histogram function.

The documentation specifies that only a list of spike trains is accepted as input, but this might be easily overlooked.
So indeed, this is an issue that should be addressed.

To resolve this, I see a couple of options:

1. Raise a TypeError if something other than a list is passed to the function.
2. Handle the case for a single spike train appropriately (as suggested by @utwrd )
3. ?

Thanks you again for bringing this to our attention! 🎉

Documentation:

• time_histogram

[Moritz-Alexander-Kern]

• decided on option 2
• extend documentation accordingly