I wanted to analyze crypto derivatives data and see if there is any seasonality/patterns to it. I have a daily dataset for about 400 days, and wanted to check for weekly/monthly/quarterly seasonality. I am new to this field, and I wanted to consult if my current approach makes sense.
Here's what I planned to do:
- Check data for stationarity using Augmented Dickey Fuller test. If the data isn't stationary, then apply log transformation and shift it by a recurring period (let's say, one week).
- Apply MA smoothing for raw data on a week/month/quarter basis and see the graphs. Apply seasonal decomposition to the resulting datasets and try to interpret these graphs.
- Use ACF/PACF on the stationary data and see if there are still significant lags remaining. If there are any, then this period is actually a recurring one.
Does this approach make sense? If there's anything missing, what would be a better way to detect seasonalities in the crypto derivatives data?