Thirty-five male mice were taken from the 32nd generation of the control
line established by El Oash et al.
(1967, Genetics, 57, 79-94) and maintained by Professor T. M. Sutherland
of the Animal Science Department, Colorado State University.
From each of these 35 litters a single male offspring was chosen
at random and weighed according to the following schedule
-
Group 1: 11 mice were weighed at ages t = 0, 3, 6, 9, 12, 15, and 18
days following birth.
-
Group 2: 10 mice were weighed at ages t = 1, 4, 7, 10, 13, 16, and 19
days following birth.
-
Group 3: 12 mice were weighed at ages t = 2, 5, 8, 11, 14, 17, and 20
days following birth.
-
Group 4: 2 mice were weighed daily from t = 0 through t = 20
days following birth.
Mice data set 1 : Id=22,...,35; t=2,5,8,11,14,17,20.
- SAS Programs and Outputs; code and output is from version 8; plots
are from version 6.
- Make the sas data set.
- Model 1 (Fixed Effect = Intercept, Day ; Random Effect = day)
- Model 2 (Fixed Effect = Intercept, Day, Day*day ; Random Effect = day)
- Model 3 (Fixed Effect = Intercept, Day, Day*day, Day*day*day ; Random Effect = day)
- Model 4 (Fixed Effect = Intercept, Day ; Random Effect = day ; UN(1) )
- Model 5 (Fixed Effect = Intercept, Day, Day*day ; Random Effect = day ; UN(1) )
- Model 6 (Fixed Effect = Intercept, Day, Day*day ; Random Effect = Intercept, day)
- Model 7 (Fixed Effect = Intercept, Day, Day*day ; AR(1) )
- Model 8 (Fixed Effect = Intercept, Day, Day*day ; ARH(1) )
- Model 9 (Fixed Effect = Intercept, Day, Day*day ; ANTE(1) )
- Model 10 (Fixed Effect = Intercept, Day, Day*day ; Unstructured)
- Report (postscript)
Mice data set 2 : Full data set.
- SAS Programs and Outputs
- Make the sas data set.
- Model 1 (Fixed Effect = Intercept, Day ; Random Effect = day)
- Model 2 (Fixed Effect = Intercept, Day, Day*day ; Random Effect = day)
- Model 3 (Fixed Effect = Intercept, Day, Day*day, Day*day*day ; Random Effect = day)
- Model 8 (Fixed Effect = Intercept, Day, Day*day ; ARH(1) )
- Model set 4 (A different way to construct the full data set.
Fixed Effect = Various Polynomials and the Unstructured Mean; Covariance
structure is usually ARH; extends the number of allowed iterations.)
Mice data set 2: Spline Models.
- SAS Programs and Outputs for a Cubic Spline model with knots at
days 3.3, 8.3 and 13.3.
- Model 1 (Fixed Effect = Cubic Spline ; Random Effect = day)
- Model 2 (Fixed Effect = Spline ; Covariance = ARH)
- Model 3 (Fixed Effect = Spline ; Random Effect = Spline; note error
in running this model.)
Mice data set 2: Splus Fit of Spline Models.
- SPlus Programs and Outputs for a Cubic Spline model with knots at
days 3.3, 8.3 and 13.3. Fixed and Random Effects are Splines. Uses LME.
Courtesy John Boscardin.
- Raw data for reading into Splus.
- Splus Code for performing the analyses. It
- Reads in the data.
- Fits the spline model with both ML and REML approaches.
- Draws spaghetti plots of fits and residuals.
- Does an eigendecomposition of the D covariance matrix.
- Fits a model with a single random effect corresponding to the largest
eigenvalue of the previous decomposition.
- Plots the population mean and the eigenvectors of the eigendecomposition.
- Unedited ML and REML Output.
- Fitted Curves (postscript).
- Residual Plot (postscript).
- Eigendecomposition (postscript).
This plots against time the population mean $X\alpha$ (divided by 10
to put it on the same scale as the other curves) (pluses), and the spline
eigenvector rotations times the square root of the eigenvalues (various
curves). This gives a picture of the population mean and the principle modes
of variation around this mean.
e-mail: robweiss at ucla.edu (replace at with @)
Rob Weiss' homepage
Repeated Measures Examples Page
Biostat 236 Repeated Measures Course Page