Trimmed means have been found to be useful in identifying underlying trends in the aggregate price data (see for example Bryan and Cecchetti (1994) and Brischetto and Richards (2007)). Notably the trimmed means have been found to produce superior forecasts of the future trend compared to traditional signal extraction tools such as core inflation. A special version of a trimmed mean is the median, which benefits from a more intuitive interpreation of being the middle item. For a general description on generating trimmed means see Dolmas (2005).
For the CPI-U inflation measure, the Cleveland FED calculates a trimmed mean inflation cutting of 16% of each side of weighted items. They also calculate a weighted median for the CPI index.
For the PCE-inflation the dallas FED calculates a trimmed mean measure cutting of the lower 31 pct and the upper 24% of weighted items.
For all indices we calculate:
Asymmetrical 24 - 31 percent trimmed mean (Dallas FED)