Scribd
Upload
824 views

Curve Estimation Explained

The document discusses how to perform a curve estimation analysis using SPSS. It explains that curve estimation allows estimating regression statistics and plots for 11 models, and is best used when the relationship between dependent and independent variables is nonlinear. It summarizes the results of linear and quadratic regression models predicting sales from advertising spending, finding the quadratic model a better fit as it accounts for diminishing returns at higher advertising levels. Residual plots also show the quadratic model fits the data better by capturing the inverted U pattern not explained by the linear model.

Uploaded by

Gerónimo Maldonado-Martínez
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
824 views

Curve Estimation Explained

The document discusses how to perform a curve estimation analysis using SPSS. It explains that curve estimation allows estimating regression statistics and plots for 11 models, and is best used when the relationship between dependent and independent variables is nonlinear. It summarizes the results of linear and quadratic regression models predicting sales from advertising spending, finding the quadratic model a better fit as it accounts for diminishing returns at higher advertising levels. Residual plots also show the quadratic model fits the data better by capturing the inverted U pattern not explained by the linear model.

Uploaded by

Gerónimo Maldonado-Martínez
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

How to Perform a Curve Estimation Analysis

The Curve Estimation procedure allows you to quickly estimate regression statistics and produce related plots for 11 different models. Curve Estimation is most appropriate when the relationship between the dependent variable(s) and the independent variable is not necessarily linear.

Select Quadratic, in additional to Linear, as

Select SAVE

Select Predicted values and Residuals in the Save

The Linear model states that the expected sales is equal to 6.584 + 1.071*advertising spending. The b1 value greater than 1 suggests that you should spend as much on advertising as you can, because you'll make that investment back and more in sales. Practically, this doesn't make much sense because you know that the market has a saturation point for advertising. The Quadratic model states that the expected sales is equal to 3.903 + 2.854*advertising spending - 0.245*squared advertising spending. The negative value for b2 means that this model suggests that past a certain point, increased advertising would actually decrease sales. More exactly, increased advertising past 2.854/(2 * 0.245) = 5.824 will decrease

The R Square statistic is a measure of the strength of association between the observed and model-predicted values of the dependent variable. The large R Square values indicate strong relationships for both models. The R Square for the Quadratic model is larger, though it is not clear whether this is due to the Quadratic model capitalizing on chance with an extra parameter .

The F, df1, df2, and Sig. columns summarize the results of the F test of model fit. The significance value of the F statistic is less than 0.05 for both models, which means that the variation explained by each model is not due to chance. The R Square statistic is a better measure of the strength of relationship.

The curve fit chart gives you a quick visual assessment of the fit of each model to the obversed values. From this plot, it appears that the Quadratic model better follows the shape of the data. In particular, the linear model seems to overestimate sales for cases with small or large values of Advertising spending and underestimate sales for cases with medium values of Advertising spending. As a further visual check, you should look at plots of the residuals versus predicted values for each

Now produce a plot showing the residuals distribution

This plot reinforces your suspicions from the curve fit plot. There is a clear "inverted U" (inverse polynomial) shape to the points, which means that there is a pattern in the data that is not captured by the

You might also like