163
Estimating Mean-Variance
Estimating mean
variance relationships
–May not be too
important for frequentist tests of the null hypothesis, because
convention often dictates the null variance we should use
•Use randomization
and/or population variances in adversarial argument
•However confidence
intervals and all Bayesian inference are statements about what
data would arise under a variety of hypotheses
•We must have some idea
about how the variance might change with the mean