[R-sig-ME] R-sig-mixed-models Digest, Vol 7, Issue 14

[Douglas Bates]

On 7/24/07, Iasonas Lamprianou <lamprianou at yahoo.com> wrote:

> thanks for the models,

My message was intended to show how you could fit the models.  I have
no plans to offer a service to fit models for others.

> but what is the conceptual difference between m1 and m2? how is the interpretation different?

Do you mean in addition to the fact that the effects for the subject
of study are fixed effects in model m1 and random effects in model m2?

If you are not familiar with the distinction between fixed effects and
random effects I suggest doing some background reading.

> would the subject with the smallest estimate be the most difficult?

Presumably.  It is your data so you would be the one to know whether a
larger score is good or bad.  The parameter estimates for the fixed
effects in model 1 represent a typical score for the subject taking
into account each student's overall ability.  In model m2 the
estimates for the random effects are relative to a "typical" or
average subject.