Mixed effect model for absolute log returns of ultra high frequency data

The influence of covariates on absolute log returns of ultra high
frequency data is analysed. Therefore we construct a mixed effect
model for the absolute log returns. The parameters are estimated in
a state space approach. To analyse the correlation in these
irregularly spaced data empirically, the variogram, known mainly
from spatial statistics, will be used. In a small simulation study
the performance of the estimators will be analysed. In the end we
apply the model to IBM trade data and analyse the influence of the
covariates.