Alan Mazankiewicz

Sure, so first what is the idea of the exponential moving average (EMA). When you have a (theoretically unbound) stream of data you are often not (only) interested in a...

About the API changes: The running exponential mean and running exponential variance require a parameter `alpha \in (0, 1)`. Therefore, the `Statistics(iterable=())` constructor would need to be modified to accept...

All right I'll implement it as another class.

Great. I was wondering: Some methods e.g. `from_state()`, `__eq__()`, `__ne__()` would be the same for both classes. Maybe it would be good to create an abstract base class that both...

all right fair enough

1. I see, you'r right. What about ExponentialMovingStatistics? If you google for that you find the concept immediately. 2. I guess I get it. You want to decay the old...

By the way, there is also a Exponentially Weighted Moving Covariance that could be implemented: https://stats.stackexchange.com/questions/6874/exponential-weighted-moving-skewness-kurtosis The post actually deals with exponentially weighted higher moments which still seems to be...

Btw. I'd suggest to change following lines in the readme: ` >>> alpha_stats = ExponentialStatistics() >>> for num in range(10): ... alpha_stats.push(num) >>> beta_stats = ExponentialStatistics(decay=0.1) >>> for num in...

For (1 to 4 ) + 6: All right! For 5: Although the post is about skewness and kurtosis what I wanted to highlight there is the **exponential covariance** which...

One thing I forgot to mention: I have not implemented the ExponentialMovingCovariance in a time-based fashion because I wanted to first get feedback from you on the how it is...