Turning noisy bathroom-scale data into a projected date: SMA vs. EMA vs. linear regression

Three common methods were tested to predict weight loss progress from daily bathroom scale readings: simple moving average (SMA), exponentially smoothed moving average (EMA), and linear regression. Each approach has its own strengths and weaknesses when it comes to forecasting when a target weight can be reached.

The SMA method calculates the average weight over the past seven days. This gives a straightforward average that falls about three days behind the current weight. However, every point in the window is given equal weight, so older data drops off quickly and the average is not very responsive to recent changes.

The EMA applies a smoothing factor of 10% to each day's weight, gradually adjusting the trend. This allows the average to move more smoothly and reduces the impact of a single outlier. However, with a smoothing factor of 0.1, the EMA lags behind the actual weight loss and responds slowly to steady progress. It's better for checking if one is still on track, but less useful for pinpointing a specific date.

Linear regression fitted a straight line through the last few data points to estimate the daily weight change. By extending this line to the goal weight, a date when the target weight will be reached can be calculated. The SMA and EMA methods don't provide a direct rate of change, while regression does. However, regression assumes a constant rate of change, which may not hold true as weight loss progresses and becomes slower. Outliers can also skew the regression line more than the other methods.