I have a set of time series data and am looking to split into different time periods to test for heteroskedascity or not over different time frames. Intially, I planned to do it this way: Take the daily change over the last month, last year, last 2 years etc. and test against eachother.
Or would the following way be better statistically? Still using daily change, split into months and test daily change over months through the last 10 years. Then repeat with years, 2 years etc.
Hope this makes sense.