Analyses of variation was performed on all data
using PROC MIXED (Littell et al., 1996) in SASR
(1996; release 6.12), using the model:
yijk ¼ l þ a½i;k þ PðkÞ þ k½i;k1 þ dj þ eijk
where: l =overall mean; a[i ,k]=fixed effect of the i-th
phosphorous level in period k; P(k)=fixed effect of
period k; k[i ,k1]=fixed carry over effect depending
on the treatment in period (k1); dj =the fixed effect
of the j-th cow; eijk =random residual error. Carryover
effects were tested for all parameters and
included in the model when they were found to be
significant (i.e., P b0.05). For all variables except
faecal P output per cow day1, Independent Normally
Distributed residuals were assumed with the same
variance for all treatments. For faecal P output per
cow day1 different variances were used for P40 to
P90 and P115 to P140, respectively. During the last
period, the cow on diet P65 was excluded due to acute
coli mastitis. No other health problems were recorded