[Non-stationary environment] Auto_regression based on which arithmetic metric?

[brrrachel]

For producing series for each <component, failure> combination, there could be different ways to do this, e.g.:

• mean value, as the average value,
• mode value, as the most common value, and
• median value, the value separating the higher half from the lower half.

The different values might have an influence on the outcoming data series. Which value would we expect to see?

[christianadriano]

I think we could start using the mean first then later on we can compare the results using the mode and median.

The code do simulate non-stationary data is here: http://web.vu.lt/mif/a.buteikis/wp-content/uploads/2019/02/02_StationaryTS_Python.html

@brrrachel could you please give a look and choose two options?