This depends on your definition or "proper".
There is no best choice for a filter width, because a filter is applied to distinguish the noise and the signal. Now imagine you've measured the sea-level near to a beach. If you are interested in the tides, you need an averaging filter over an hour. To measure the distribution of wave-heigths, a filter over seconds is useful. And if you want to get the influence of the climate, the filter must span some years.
Therefore you need a physical motivation to choose a filter parameter ever and the variance is not a sufficient base.