I work with cell cultures. Within each culture, cells are plated on to coverslips which are then put on a microscope so that measurements can be made on typically 30 cells at a time (this number varying between coverslips). I measure a quantity which I call DCF and want to see whether it varies according to a treatment (EGTA) with two levels (Y or N, i.e., yes or no). The treatment has to be applied to all the cells on a coverslip. I use 3-4 cell cultures and, within each culture I use 3-4 coverslips for each level of EGTA (i.e., 6-8 coverslips per culture).
In this design, CULTURE and COVERSLIP are random effects and EGTA is a fixed effect. COVERSLIP is nested within EGTA.
I guess this counts as complete randomised block design with the blocks being cultures and the plots being coverslips. It also has repeated measures within each coverslip although unlike most examples of repeated measures these are not repeated measures in time or in position.
I can use PROC MIXED as follows:
proc mixed data=""> class CULTURE COVERSLIP EGTA;
model DCF=EGTA /ddfm=satterth;
random COVERSLIP CULTURE;
run;
In my data table I have labelled the coverslps differently within each culture so I assume that I do not need to specify nesting explitly. Anyhow, running the above with COVERSLIP(EGTA) instead of COVERSLIP as a random factor produces the same result.
My problem is that from eyeballing the data I suspect that the effect of EGTA may vary significantly between cultures. How do I test for this since EGTA is a fixed term and cultures is a random term.
Including the interaction term in the model i.e. DCF=EGTA EGTA*CULTURE results in the log warning:
WARNING: Did not converge
Including the interaction term in the random statement i.e. random COVERSIP CULTURE CULTURE*EGTA results in the log note:
NOTE: Estimated G matrix is not positive definite.
and an F value for the EGTA term that is the same as that obtained without the interaction term.
Can anyone help?
Ron
Centre for Cardiovascular Biology and Medicine
New Hunt's House (room 2.32A)
GKT School of Biomedical Sciences
King's College London
Guy's Campus
London SE1 1UL
Tel 020 7848 6215
Fax 020 7848 6220
