Timeline for Understanding spline transformation and regression coefficients
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 20 at 2:38 | comment | added | Shawn Hemelstrand | I also don't understand this distinction and don't believe it to be accurate. I have seen splines used for main effects for many people who are experts in this area. Furthermore, I don't think this answer actually elucidates the core question asked by OP. The answer so far more shows how one can program the spline without a clear description of what is going on under the hood, which is likely what OP was asking for. | |
| Feb 16 at 14:47 | comment | added | rolando2 | @Roland I've never heard that distinction made before (not that I am so knowledgeable about splines). I am curious why you make it. I would think that what's good for estimating a main effect of interest would be good for estimating a confounder's effect and vice versa. | |
| Feb 16 at 6:36 | comment | added | Roland | @rolando2 I'm not sure what you want me to elaborate. I might use a spline to account for the nonlinear effect of a nuisance variable. I wouldn't use a spline for the main effect of interest. | |
| Feb 13 at 16:07 | comment | added | rolando2 | @Roland Could you elaborate on "not that interested in the coefficients"? | |
| Feb 13 at 13:54 | comment | added | denis | ok, but still it does not answer my question: Why and how this transformation produce the same prediction as my understandable piecewise regression ? What is the rationale of this tranform ? Where does it come from ? How can I reverse it to the slope of each first degree polynomial ? | |
| Feb 13 at 8:12 | comment | added | Roland | Generally, when you use splines, you are not that interested in the coefficients ... | |
| Feb 13 at 7:39 | comment | added | denis | thank you for the transformation. But I do not understand the rational of this, nor how to interpret the coefficients when you use such transformation in a regression | |
| Feb 9 at 10:10 | history | answered | Roland | CC BY-SA 4.0 |