Timeline for Linear spline and 'interaction' p value
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 28, 2023 at 1:34 | comment | added | David B |
age + age_slope_change is the same as (age-25):age_bin + age. If you recode age_bin as 0 and 1 then (age-25):age_bin is literally the same as age_slope_change.
|
|
| Mar 27, 2023 at 20:30 | vote | accept | LucaS | ||
| Mar 27, 2023 at 20:30 | comment | added | LucaS |
Thanks - I will accept your answer as it's the closest match to my question. I still couldn't get those two parameterisations to match up though - age + age_slope_change and (age-25)*age_bin.
|
|
| Mar 27, 2023 at 15:04 | comment | added | David B | Actually no! They're similar, but there's a very important difference. See the edit to my original answer. | |
| Mar 27, 2023 at 15:04 | history | edited | David B | CC BY-SA 4.0 |
added 825 characters in body
|
| Mar 26, 2023 at 21:53 | comment | added | LucaS | Thanks @DavidB. This is probably the closest to helping me understand my problem. It's a different parameterisation. What I don't still fully get is why the estimates from the two models aren't the same (they are close but not exactly the same). Isn't a model with age + age_slope_change (where age_slope_change = 0 before chngpt and (age-chngpt) after), essentially the same as a model with age * age_bin (where age_bin = 0 before chngpt and 1 after)? | |
| Mar 24, 2023 at 15:01 | history | answered | David B | CC BY-SA 4.0 |