Page 1 of 1

[SOLVED] design vector at level 2 is the wrong length

Posted: Sat Feb 08, 2014 7:56 pm
by kaiserdominici
*SOLVED*

It was indeed the missing data. Once I listwise-deleted them and tried again building up from an empty model, it worked. Original post below.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------

Hi all,

I was trying to reproduce the procedure in Chapter 5 in the Manual Supplement to MLwiN (http://www.bristol.ac.uk/cmm/software/m ... nt-web.pdf) using my dataset.

I get the error in the Subject when I run the command "star" to start the model. I read in a previous post that this may be due to missing data, and indeed my model uses 259 out 320 cases. Some info on the model:

Code: Select all

y_ij = β_0j + β_1j x t_ij + β_2 x t_ij^2 + ε_ij
β_0j = β_0 + υ_0j
β_1j = β_1 + υ_1j
Where y is a continuous variable, i are measurement occasions and j is an individual, and t is a time variable that goes from 0 to 4, and it has also been squared to investigate acceleration in growth. Since not all individuals took all measurements, there are some missing values at level 1. Since it is a repeated measures dataset, I wanted to try and allow for AR(1), hence my reference to that chapter in the manual.

Many thanks for any help :)

k.