Hi all,
I have found quite a prominent U shaped curve between two of my variables that I would like to test in SPSS, to assess whether this is significant.
I have conducted a curve estimation using a quadratic model but I wanted to make sure that this is the best way to test this? What is the difference between this and a non-linear regression?
Thanks!
Hi Gaby,
Your approach sounds reasonable, i.e., constructing a regression model with the predictor and predictor^2 terms. You can compare the explained variance for this model relative to one with only the linear effect of predictor to see how it performs. As well, you could compare a scatterplot of the residuals to see whether the quadratic model better accounts for the non-linear association.
You can think of this approach as a type of non-linear regression, which is a broad term.
Adding a quadratic term to the model is a common approach, but it is flawed. See the below Data Colada posts, which describe the problem and outline a better approach (the second also includes a link to a published paper on this):
http://datacolada.org/27
http://datacolada.org/62
Quadratic terms (^2) in regression equation after corresponding linear term.
Hope this helps
Matt
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS