'pychattr' library in Python, 'n_simulations' parameter

Question

Does anyone know if it is possible to use n_simulation = None in 'MarkovModel' algorithm in 'pychhatr' library in Python?

It throws me an error it must be an integer, but in docsting i have information like that:

'n_simulations : one of {int, None}; default=10000'

I`d like to do something like nsim = NULL in 'markov_model' in 'ChannelAttribution' package in R, these two algorithms are similarly implemented.

I don`t know how does it works exactly, how many simulations from a transition matrix I have using NULL.

Could anyone help with this case?

Regards, Sylwia

Answer 1

I added an issue:

https://github.com/jmwoloso/pychattr/issues/10

A code doesn`t support None value despite a documentation.

https://github.com/jmwoloso/pychattr/blob/70ec07c05fc555afcd3683b258894463f34a40ae/pychattr/channel_attribution/_markov.py#L424

Answer 2

Out of curiosity I spent some minutes staring intensely at the source code of both pychattr module and ChannelAttribution package.

I'm not really familiar with the model, but are you really able to call this in R with "nsim=NULL"? Unless I missed something if you omit this parameter it will use value 100000 as the default and if parameter exists, the R wrapper will complain if it's not a positive number.

Regards, Maciej

Answer 3

I checked that 'pychattr' (Python) doesn`t support value None but it supports n_simulations = 0 and it sets n_simulations to 1e6 (1 000 000).

'ChannelAttribution' (R) replaces nsim = NULL and nsim = 0 to nsim = 1e6 (1 000 000) too.

In latest version of 'ChannelAttribution' (27.07.2020) we have nsim_start parameter instead of nsim and it doesn`t support 0 or NULL value anymore.

Important: default value of nsim_start is 1e5 (100 000) and from my experience it`s not enough in many cases.

Regards, Sylwia