Jon Heron posted on Friday, May 13, 2011 - 2:52 pm
I am fitting a growth model with a baseline and 3 linear splines to 8 data points
To illustrate the univariable effect of a few predictors for a up-coming presentation I am using the "Adjusted estimated means" option.
I have a binary predictor as the only covariate in my model so am using the "name covariate set" option to create two estimated lines corresponding to predictor=0 and predictor=1.
This yields a graph with two lines as I would expect, which roughly agrees with the conclusions based on the model parameter estimates, however the vertical distance between the two lines in my plot at baseline is greater than the estimated effect of my predictor on the intercept.
Am I perhaps using the graphing option inappropriately?
Jon Heron posted on Monday, May 16, 2011 - 11:49 am
I inferred from your response that I had probably messed up somewhere along the line so went back and checked everything again.
I had been attempting to move my intercept term from the baseline position to the end-point and also to the two elbow joints of my 3-spline model. Working out how to tweak all the loadings is a little tough on the brain, particularly on a friday afternoon.