These are data from a one year study of gum disease conducted in the mid 1990s.
The measurements being followed over time are for whole-mouth average pocket
depth and whole mouth average attachment loss. Visits were at baseline, 12 weeks,
24 weeks, 36 weeks, and 52 weeks. Pocket depth and attachment loss were measured
at many sites within each participant's mouth and then averaged (at each visit).
The variable called sites gives the number of sites used in the averages.
Pocket depth and attachment loss are both measures of how far the gums have pulled 
away from the teeth, hence smaller values are better.  Whole-mouth average
is used because the measurements within mouth are highly correlated in a 
complicated way which is not yet fully understood by dental researchers.

Covariates available include race, gender, age, smoking status (yes/no), and a 
treatment. The treatment had five levels: no treatment and four levels of an 
active substance in a gel. Participants were asked to rub the gel on their gums
twice daily. The lowest level was a placebo (another control group, like the 
"no treatment" group). The remaining three levels were low, medium, and high 
concentrations of the active ingredient in the gel.

The patients were recruited via a single midwestern dental research clinic.
There were 130 participants.  Missing values are denoted by a period.

NOTE: The data set dental.s.dat has all missing values changed to the 
S-plus needed "NA" instead of a period.