Polynomial Trend Line Equation

[dodger7]

Hi Folks

I have created an order 2 polynomial trend line from my data (sales volumes from mar-10 to jan-11).

The equation shown on the chart is:

y = 61.578089x2 - 163416.228904x + 108425159.760839

Does anyone know how to get the values for the trend line using the above formula?

Thanks in advance

 

[VoG]

With an x value in A1 try

=61.578089*A1^2 - 163416.228904*A1 + 108425159.760839

 

[dodger7]

With an x value in A1 try

=61.578089*A1^2 - 163416.228904*A1 + 108425159.760839

Hi Peter,

Is a known x value a sale? Sorry I am new to this kind of statistical forecasting

Thanks
Jamie

 

[VoG]

x is whatever you plotted on the x-axis of the chart.

 

[dodger7]

I had guessed that, but x is dates. Will this make sense to the forumla, to start with an excel date value?

I tried it as an example but the result was millions out

 

[VoG]

Sorry, I'm not sure. It does generate rather large numbers. But a date in Excel is just the number of days since 1 Jan 1900.

 

[dodger7]

Sorry, I'm not sure. It does generate rather large numbers. But a date in Excel is just the number of days since 1 Jan 1900.

Yes it's strange... The numbers do seem to increase in the way i'd expect, just at a few hundered million out from the scale I was expecting!

Thanks for your help anyway and time

 

[shg]

If you have a line graph, then the chart considers the x values to be categories (1,2,3,...}, and the dates just to be the names of those categories. So 1 corresponds to the x in the first date on the chart, 2 to the second, ...

If you change the graph to a scatter plot, it will use the actual dates (assuming they are dates, not text that looks like dates), and the coefficients will change dramatically.

 

[dodger7]

If you have a line graph, then the chart considers the x values to be categories (1,2,3,...}, and the dates just to be the names of those categories. So 1 corresponds to the x in the first date on the chart, 2 to the second, ...

If you change the graph to a scatter plot, it will use the actual dates (assuming they are dates, not text that looks like dates), and the coefficients will change dramatically.

SHG, thanks for your input - this now makes sense. By me using date vales I was multiplying to the power of thousands rather than just the index of the variable.

Thanks to both of you for your time