Trendline s. regression analysis

[Kris]

Using the same set of data, I created a scatter diagram
with trendline, equation, and R-squared, and performed a
linear regression analysis in Excel. Both methods
produced the same coefficient and R-squared, but they
resulted in different intercepts. Regression analysis
with SPSS yields the same results as the Excel regression
analysis. Why would the intercept as determined by the
trendline be different?

 

[Michael R Middleton]

Kris -

Using the same set of data, I created a scatter diagram with trendline,

equation, and R-squared, and performed a linear regression analysis in
Excel. Both methods produced the same coefficient and R-squared, but they
resulted in different intercepts. Regression analysis with SPSS yields the
same results as the Excel regression analysis. Why would the intercept as
determined by the trendline be different? <

Maybe your chart was a Line chart type, instead of an XY (Scatter) chart
type.

The trendline on a Line chart uses x values 1,2,3,...

The trendline on an XY (Scatter) chart uses your actual x values.

- Mike Middleton, www.usfca.edu/~middleton

 

[Kris]

Thanks for your help. However, it is definitely a scatter
(XY) diagram with a linear trendline. The Y-intercept has
not been forced to zero or any other value.

 

[Harlan Grove]

Using the same set of data, I created a scatter diagram
with trendline, equation, and R-squared, and performed a
linear regression analysis in Excel. Both methods
produced the same coefficient and R-squared, but they
resulted in different intercepts. Regression analysis
with SPSS yields the same results as the Excel regression
analysis. Why would the intercept as determined by the
trendline be different?

What was the magnitude of the difference in the intercepts?

FWIW, Excel's chart trendlines are more reliable than either the Regression tool
in tools > Data Analysis... or the LINEST worksheet function because the former
is based on a more numerically robust algorithm, so it's more likely to be
correct - even if it doesn't agree with SPSS.

 

[Jerry W. Lewis]

I have not compared to SPSS, but the chart trendline is numerically
better than either S-PLUS or SAS.

Jerry