IMHO, part of the problem is with your assumption that the trendline for the non-cumulative and cumulative data is a 3-degree polynomial.Īrguably, it is true for your cumulative data. But at the risk of beating a dead horse, I would like to offer the following for consideration. Again, compare 'estimated y' with both sets of coefficients to see if the difference matters. Then enter the difference between the next dates and, resulting x-axis values of 1, 31, 60, 91, etc.īut even then, the trendline and LINEST coefficients are different. Those are very large numbers for example, is 42370. I prefer to use Scientific with 14 decimal places that works regardless of the magnitude of the coefficients.Īnother potential problem: your x-axis values are dates. The difference does not matter as long as both sets of coefficients result in similar 'estimated y'.īut remember: to use the trendline coefficients, it is important to format them with sufficient precision. Some reasons are: (a) different algorithms and (b) differences in the arithmetic implementation (order of operations). Nevertheless, you might still see a difference. because the actual x-axis 'values' are just labels.Īfter changing the chart type to XY Scatter, delete the trendline and re-add the trendline to get correct coefficients. The point is: in Line charts, x-axis values used for the trendline are always 1,2. Although much of the information there is incorrect(!), the 'Note' is mostly correct.

You must use an XY Scatter chart in order to get correct trendline coefficients.

The first and primary problem is: you are using a Line chart.

Excel polynomial trendline equation