Weighting Linear Trend Lines

[Rob P.]

I have historic data which is on a line graph to which, I add a trendline in order to measure the accuracy of my forecast. I would like to have the ability to weight the near term dates (say within the last year) heavier with the premise that near term results are a closer approximation of the true trend rather than data from 2 or 3 years ago.

I've seen a post referring to using bubble charts but I wasn't able to see the impact of the weighting when I changed the bubble sizes.

Any ideas (or external software) which can help solve this problem?

Rob P.

 

[Oaktree]

Not sure I understand. It sounds like you've developed a forecast, which you've backtested and plotted against the actuals from prior years. Where does the weighting come in? Wouldn't you just show that the gap between actual and forecast is small for recent years but perhaps larger for years far in the past?

Or, is your question how to actually develop the forecast so that it weighs recent years' input more? If so, I'm a bit confused by your use of linear trend lines... perhaps you're looking for exponential smoothing (part of the analysis toolpak)?