Allow option to mitigate NaNs and 0.0s at the top of returned dataframe

[bukosabino]

Allow option to mitigate NaNs and 0.0s at the top of returned dataframe

[bukosabino]

Hi @cobrobrown ,

Absolutely, you are welcome to contribute. You can start doing a Fork in GitHub of this repository, then you can work in your own repository. When you finish, you can do a Pull Request. Then, I will check your pull request and probably I will merge.

Let me know if you need anything more!

Best,
Darío.

[bukosabino]

Added on new 0.4.1 version

[RichardDale]

I'm requesting this issue be re-opened.

Further to this, I think this should be optional for every indicator and not mandatory.

For example, in normal use on most technical analysis/charting app, a 20 bar indicator doesn't start until bar number 20. With the min_periods=0, values are populated for the first 19 bars.

Source: I've used plenty of TA apps over the past 20 years. MetaStock, AmiBroker, various charting libraries, TradeStation, Multicharts, NinjaTrader, RightEdge, Advanced GET, Fcharts, Market Analyst, Omnitrader to name a few... None of them plot anything in the first 19 bars on a 20 bar indicator.

[bukosabino]

Hi @RichardDale and @cobrobrown ,

I am agree with @RichardDale , I think we should delete the min_periods=0 and fill_value=x.

Let me know if you are agree.

Best,
Dario

[mementum]

What @RichardDale describes is what can be described as a basic guarantee, i.e.:

• The length of the input buffer is equal to the length of the output buffer
• As long as the minimum period of the indicator has not been met, the value to be produced has to be a non-value

The standard fill value for non-defined output values is NaN, for a very simple reason:

• It is NOT a value and defined so in a standard and it can be tested for such (which allows for easy filtering, hence the ...na machinery in pandas and other frameworks
• It is skipped by plotting frameworks like matplotilb