jointly.calculate_magnitude returns 0 for NaN inputs

[AlexS12]

• jointly version: master
• Python version: 3.9.7
• Operating System: Ubuntu (inside Windows Subsystem for Linux)

Description

In the usage example, faros_df lacks acceleration data in some rows (x, y, z components are NaN). However, returned magnitude for that rows is 0.0 and I would expect NaN. Is there a reason for that?

What I Did

faros_df["Accel Mag"] = jointly.calculate_magnitude(
    faros_df, ["Accel X", "Accel Y", "Accel Z"]
)

Results in

Accel X	Accel Y	Accel Z	ECG	Accel Mag
-88.0	771.0	-531.5	-21.0	940.572831
NaN	NaN	NaN	-10.0	0.000000
-86.0	779.0	-539.5	NaN	951.471098
NaN	NaN	NaN	-2.0	0.000000
-82.5	781.0	-543.0	NaN	954.785971
...	...	...	...	...
NaN	NaN	NaN	-9.0	0.000000
-166.5	896.0	-316.5	NaN	964.733383
NaN	NaN	NaN	12.0	0.000000
-156.5	900.0	-320.5	NaN	968.097361
NaN	NaN	NaN	28.0	0.000000

7400 rows × 5 columns

Possible solution

According to DataFrame.sum method documetation (https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.sum.html)

By default, the sum of an empty or all-NA Series is 0.

pd.Series([], dtype="float64").sum()  # min_count=0 is the default
0.0

This can be controlled with the min_count parameter. For example, if you’d like the sum of an empty series to be NaN, pass min_count=1.

pd.Series([], dtype="float64").sum(min_count=1)
nan

Thus, I would modify np.sqrt(np.square(data).sum(axis=1)) in calculate_magnitude by np.sqrt(np.square(data).sum(axis=1, min_count=1)) or use np.linalg.norm(data, axis=1).