stdev does not take ddof = 0

[dnabb]

Which version are you running? The lastest version is on Github. Pip is for major releases.
0.3.2b0

Describe the bug
Calling stdev with ddof=0 (and all other derived indicator, like bbands) actually returns a standard deviation with ddof = 1.

To Reproduce

import pandas as pd
import pandas_ta as ta

data = {'close': ['1', '2', '2', '1']}
df = pd.DataFrame.from_dict(data)

df.ta.stdev(length=4, ddof=0)[3] # Shoukd be 0.5

df.ta.stdev(length=4, ddof=0).equals(df.ta.stdev(length=4, ddof=1)) # Should be False

Expected behavior
Should be the same as:

df.rolling(window=4).std(ddof=0)['close']

Additional context
The issue seems to be this validation which fails with ddof = 0.

[twopirllc]

Hello @dnabb,

Doh! Nice catch!

In the mean time, edit your local copy of variance and stdev and replace that validation with the following until the code is updated.

ddof = int(ddof) if isinstance(ddof, int) and ddof >= 0 and ddof < length else 1

Hope this helps!

Kind Regards,
KJ

[twopirllc]

@dnabb,

Also I have added the changes to the development branch if you want to test it out.

$ pip install -U git+https://github.com/twopirllc/pandas-ta.git@development

Please let me know if it works as intended if you try the development branch.

Thanks,
KJ

[twopirllc]

Hello @dnabb,

New stable version v0.3.14b should help resolve this Issue. Please let me know otherwise.

$ pip install pandas_ta

Thanks,
KJ

[twopirllc]

Hello @dnabb,

I assume by no response that the solution provided was sufficient. Thus I will be closing this issue in a few days.

Kind Regards,
KJ

[twopirllc]

Hello @dnabb,

This should be fixed now. If you can test it out on the development branch and let me know, that would be great.

$ pip install -U git+https://github.com/twopirllc/pandas-ta.git@development

Thanks
KJ

[dnabb]

Hello @twopirllc ,

I can confirm the issue seems to be solved in 0.3.18b0.

[krazykoder]

Just reopening this post. I observed TALIB STDDEV and pandas_ta std_dev generates different output. A reference to this issue is noted here as well TA-Lib/ta-lib-python#69 (comment). Though not sure this is exactly the same or not.

But I just wanted to confirm with others that pandas_ta standard deviation should be safe to use right since it matches my final indicator calculations cross referencing tradingview. I will stop using talib for this function.

code to try below

length=100, fast=50, slow=50

import talib as talib
talib.STDDEV(closeSeries, timeperiod=fast)

2022-06-08 12:00:00    12.704822
2022-06-09 08:00:00    12.060018
2022-06-09 12:00:00    10.837907
2022-06-10 08:00:00    10.635564
2022-06-10 12:00:00    10.512141
Length: 2038, dtype: float64

import pandas_ta as ta

ta.stdev(closeSeries, fast)

2022-06-08 12:00:00    12.833808
2022-06-09 08:00:00    12.182457
2022-06-09 12:00:00    10.947939
2022-06-10 08:00:00    10.743542
2022-06-10 12:00:00    10.618866
Name: STDEV_50, Length: 2038, dtype: float64

[krazykoder]

Closing this issue, I found the answer in the same talib post (did not read fully sorry!)

While using TALIB need to add nbdev parameter which is np.sqrt(length / (length - 1)

talib.STDDEV(closeSeries * 2, timeperiod=fast, nbdev=np.sqrt(fast / (fast - 1))