Questions on sliding window analysis.

[andrewcz]

In summary i am trying to answer the following questions -
Is there a lag consistent/persistent through time
What is the strength of lag
What is time delay of the lag

So my thought is that the sliding window analysis is for understanding persistence in causal effects.

My question is what is the difference between -

window_step=1
window_length=7

If I have a set of financial time series described by the below code. I would think that the lead lag happens quite quickly hence the short window step (?). the below is my starting code

import pandas as pd
import yfinance as yf


df = yf.download(['AAPL','GOOG','MSFT','DIS', '^GSPC'], start="2021-01-01")['Adj Close'].dropna()
dataframe=df.pct_change().dropna()
data=dataframe.to_numpy()
data.shape
var_names = ['AAPL','GOOG','MSFT','DIS', '^GSPC']
dataframe = pp.DataFrame(data, var_names=var_names)
tp.plot_timeseries(dataframe); plt.show()

N = 5
T = data.shape[0]
datatime = np.arange(T)

[andrewcz]

The visual get from running the code -

Please assist with comments or suggestions @jakobrunge

[jakobrunge]

window_step is the stepsize in units of the time series data resolution of the windows over which each causal graph is estimated.

[jakobrunge]

window_step and window_size have not much to do with time-lags, it's just that tau_max should be much smaller than window_size, otherwise you have no samples left to estimate anything on.