Let's say we have a universe of 100 stocks, S1, S2, ... S100, and their historical price data (daily, for the past 50 days). What would be the most robust/widely used method to capture the price movement of the universe as a whole over the course of the history? Any de-facto approach in the quants/finance field?
My thought is that this boils down to the choice of what moment for each probability distribution at timestep t, as each data point should exibit the universe's price tendency.
My simplest approach would be to
- detect outliers with X percentile, and remove them
- calculate mean of the universe for each timestep t
- plot the mean as a line graph