# [R] simple ANOVA problem

Kyriakos Kachrimanis kgk at pharm.auth.gr
Sun Jun 30 19:34:18 CEST 2002

```Hello list members,

this is rather a simple question but I can't find the solution myself, so I
would appreciate your help and I apologise in advance for my ignorance.

I have a data frame with a header containing variable names and I have 4
independent variables and 5 dependent variables. It looks like this:
x1     x2  x3     x4     y1     y2     y3     y4     y5
1     1     33     55     4       0.3     0.38  1.55     30.1
1     1     52     56     5.5    0.28   0.36  1.51     20.8
1     1     72     55     7.9    0.27   0.36  1.52     15.1
1     2     33     34     4.7    0.34   0.41  1.55     23.2
..............
and it contains a few more data points. I want to estimate the significance
of the independent variables on each of the dependent but not all the
independent variables are  random factors. x1 is a category (qualitative)
taking 2 values, x2 and x3 are quantitative fixed-effects factors with 3
levels and x4 is a random factor highly correlated with x3 (so for
simplicity, I am thinking of skipping one of these 2 factors from the
analysis). Furthermore, there is no replication, so I have to estimate the
experimental error in terms of the higher-order interactions.

I tried the aov( ) function with a model like y2~x1*x2*x3+Error(x1:x2:x3).
The result I got was different from what the Statistica (ANOVA with
fixed-effects model and customised error term) is producing. I suspect that
the difference arises from the fact that I have to define somehow the nature
of the dependent variables otherwise they are treated as random quantitative
factors. I couldn't find a relevant example in the manuals. Can somebody
provide some help?