0

I am going to be using the logistic regression in which I will use L2 Regularization. I have these 4 rolling standard deviation variables. Here are the results of the Augmented Dickey-Fuller Test for stationarity. It says they are stationary according to the p-values which are all below 0.05:

(-18.93610985313199, 0.0, 21, 21551, {'1%': -3.430653469865324, '5%': -2.8616741235464906, '10%': -2.5668413902065788})

(-14.904236674495897, 1.491198925557711e-27, 43, 21520, {'1%': -3.4306539070720308, '5%': -2.861674316767364, '10%': -2.5668414930543557})

(-15.369581186780854, 3.5231827878994372e-28, 44, 21459, {'1%': -3.430654771070804, '5%': -2.8616746986063397, '10%': -2.5668416962999574})

(-4.289075846312884, 0.000463964612151969, 15, 21272, {'1%': -3.4306574506058665, '5%': -2.861675882809773, '10%': -2.5668423266289735})

I want to standardize these 4 variables for L2 reg according to this point made here: https://stats.stackexchange.com/a/195391/363734

Here is a before and after standardization, it seems there isn't much of change even though we mean centre around zero. The question I would like to know is, is it better to difference these features before standardizing or should I just keep them as they are and then standardize?: Before Scaling

After Scaling

DomIsAwesomee
  • 5
  • 2
  • Can you clarify what you mean when you say you're going to "difference" these features before standardizing? – Kenneth Granahan Mar 16 '23 at 09:07
  • Difference as in make the series stationary (compute the differences between consecutive observations). Even though according to the ADF test it says the raw rolling standard deviations are stationary, it does not look stationary, e.g. a white noise process. From experience, is it okay to feed the standardized version of the rolling standard deviations into a regularized logistic regression or should I difference it first and then standardize it? – DomIsAwesomee Mar 16 '23 at 14:10

0 Answers0