[R-sig-ME] how do I describe heteroscedacity

Lize van der Merwe lize@t@t@ @ending from gm@il@com
Fri Aug 24 09:50:54 CEST 2018

I  used lme to fit a model which includes different variances for operators.

My question is:

What is the best way to report the operator standard deviations?   

Here are the details.

My data contains an outcome, with predictors load (4 different loads) and
operator (10 different operators).  It is a very large unbalanced data set,
with some operators only working on a single load while others operated on
all 4 loads.  I am interested in the load means, the operator means and the
operator variances.  The latter is my problem.

I used the model  

lme(Outcome~Load,random=~1|Operator,weights = varIdent(form= ~ 1 |

I used intervals() to obtain estimates with 95% CIs of the effects of the
loads and the operators.   

Because of identifiability, the model yields delta, the ratio of specific
operator standard deviation to first operator standard deviation with their
95% CIs.  The attached plot shows the deltas, with their confidence

Is it correct to state that, because the some of the CIs do not overlap, the
observer 4 is clearly less precise than some other observers? 

Thankyou in advance for advice.


Lize van der Merwe 



-------------- next part --------------
A non-text attachment was scrubbed...
Name: Observers.pdf
Type: application/pdf
Size: 2121 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20180824/b848eb45/attachment.pdf>

More information about the R-sig-mixed-models mailing list