www.cmm.bristol.ac.uk/forum

Dear MLwiNners,

I have a small question regarding the prefered notation concerning the notation of the estimated β coefficients reported in MLwiN when including the estimates in the result section of an article.
When reporting the specific coefficients in the text, should one use the notation β (as MlWin calculates estimates for the population) or b (as these estimates are based on the sample values)? (e.g. something like β=1.23, p<.05)
What is the most common notation? Does anyone have examples in papers or manuals?
And does this differ from reporting single-level regression results?

Kind regards,
Rasul

Professor Kelvyn Jones has provided the following response:

The fixed effects, the βs, represent the ‘population’ means and these are similar to the parameters of a single-level regression model. The estimates may be portrayed by a non-Greek B or by placing a hat over the parameter, β̂. There would appear to be no general accepted convention but differences occur by discipline and journal. The best approach is to choose a journal and have a look at previously published papers – the Multilevel Gallery (http://www.bristol.ac.uk/cmm/research/gallery/) may be of help here.