nonlinear regression

[kc011c]

Hi all. I've got two sets of data, (let's say, one series = $ sales, the other is number of salespersons). Assuming a diminishing marginal return relationship (i.e. one more salesperson will bring in less in sales than the last salesperson did), I want to do a non-linear regression that will allow me to see what is the optimal number of salespeople (zero marginal returns). I can select "regression" from Data Analysis, but the summary statistics are all from a linear regression. I can only change the trendline in the chart to a polynomial regression. I need to know how to get the non-linear output as well as how to evaluate the regression model (do I still look at the R^2 value from the trendline?). Thanks

 

[sann0638]

Hi there,
if you take logs of the data, then do a linear regression, it will be analogous to doing a non-linear regression (I think) with the coefficient of the log of the parameter being the power of the parameter. The R^2 can then be used as normal.

 

[kc011c]

Hi there, thanks for the reply. But how can I see where the optimal number of salepeople is with a linear regression? With the curved trendline I can see where it peaks... and perhaps use solver to tell me where that point is...Appreciate any light you can shed on this matter.

 

[sann0638]

Hello again,
the linear regression will give you the best fit line, ie if your best fit is:

log(sales)=2.5*log(people) then your formula that you need to max is sales = people^2.5

(not the best example because this doesn't max, but hopefully this helps)

Edit: Not very well explained - if you send me an email I can work you up a solution really quickly.
This message was edited by sann0638 on 2002-10-18 06:17