This folder includes two datasets.  

One dataset ("Smaller dataset") includes one visit for one patient;  that patient's 
attachment loss data is plotted in Figure 14.1 and analyzed in Chapter 15.  

The second dataset ("Larger dataset") includes one visit for each of 130 patients.  The 
single patient in the smaller dataset (above) was drawn from this larger dataset, as were 
the patients analyzed in Reich, Hodges, and Carlin (JASA 2007) and Reich & Hodges 
(Biometrics 2008).  

More details follow.


**Smaller dataset (one patient)**

The file "Periodontal_dataset_Fig14p1" has 168 rows, one for each measurement site (6 
sites per tooth, 28 teeth with periodontal measurements;  if the third molars ["wisdom 
teeth"] were present, they were not measured).  The sort-order of rows is discussed below.  
The 12 columns are:

-- Patient ID:  This patient is in the larger dataset described below;  this is his ID 
number in that larger study.

-- Arch:  Upper (Maxilla) or lower (Mandible)

-- Tooth:  Tooth label using the World Dental Federation notation (http://
en.wikipedia.org/wiki/FDI_World_Dental_Federation_notation).  Each tooth's label has the 
form XY, where X indicates the quadrant and Y the tooth in the quadrant.  The quadrants 
are 

    1:  Upper right
    2:  Upper left
    3:  Lower left  (note!  They circle around the mouth)
    4:  Lower right

The teeth are numbered from the front/center toward the rear:

    1:  Central incisor
    2:  Lateral incisor
    3:  Canine
    4:  First premolar (aka first bicuspid)
    5:  Second premolar
    6:  First molar
    7:  Second molar
    8:  Third molar (aka wisdom tooth;  not included in this dataset)

Thus the tooth labelled 36 is the lower left first molar.

-- side:  Inner or outer side of the tooth.  Values are:  Lingual (tongue or inner side) 
and Buccal (cheek or outer side)

-- site:  Measurement location on the arch, tooth, and site.  Values are Distal (farthest 
from the front/center of the mouth), Direct (in the middle of the tooth), and Mesial 
(closest to the front/center of the mouth).  

-- Arch sort, Side sort w/in arch, Tooth sort w/in arch, site sort w/in tooth/arch:  These 
variables are used to sort the observations in the order needed for computing neighbor 
pairs.  The sort orders are:

   -- Maxilla, then mandible
   -- Lingual, then buccal
   -- tooth:  In this file, teeth are sorted from right to left within the maxilla
              then from right to left within the mandible.  Thus the order is 
              17, 16, ..., 11, 21, 22, ..., 27, 47, 46, ..., 41, 31, ..., 37. 
   -- site:  The sort order for sites within teeth is Distal-Direct-Mesial for teeth on the 
             right side of the mouth and Mesial-Direct-Distal for teeth on the left side
             of the mouth.

-- site no:  Sites are numbered 1 to 168;  these are the same as the index (row number) 
values used in the R code to construct the adjacency matrix Q (proportional to the ICAR's 
precision matrix).  

-- al1:  Attachment loss measurement at that site (distance from the cemento-enamel 
junction to the bottom of the periodontal pocket)

-- pd1:  Pocket depth measurement at that site (distance from the free gingival margin to 
the bottom of the periodontal pocket).


The file "Periodontal_dataset_Fig14p1_R_code.txt" includes R code that constructs the 
adjacency matrix Q (proportional to the ICAR's precision matrix) for this dataset;  this Q 
was used in the analyses discussed in Chapters 14 and 15.  This code uses the sort order 
described above;  sorting on "site no" produces the right order.  


If you plot these data, you'll note that they don't look quite like Fig 14.1, which is 
taken from Reich & Hodges (JSPI 2008).  In particular, the labeling of sides (buccal/
lingual) and sort order within arch used to draw this plot are not the same as the side 
labeling and sort order in the present dataset.  I think -- but I wouldn't bet my life on 
this -- that the sort order in the present dataset is correct.  These discrepancies have no 
effect on the analyses presented, which are symmetric the buccal and lingual sides and the 
right-to-left sort order.  These discrepancies would matter in other analyses, e.g., that 
included buccal vs. lingual as a fixed effect.  


** Larger dataset **

The file "Periodontal_dataset_all_visit1" has 21,840 rows, 168 rows for each of 130 
persons.  This dataset includes Visit 1 data for each of 130 patients and is extracted from 
a still larger clinical trial dataset that included up to 5 visits for each subject.  The 
smaller dataset, above, is subject ID 106 in the present dataset.  Some of these data were 
analyzed in Reich, Hodges, and Carlin (JASA 2007).  In that paper, the patient in the one-
patient plots is ID 101, and the patient called "Patient 15" is ID 115.  For the analyses 
in Reich, Hodges, and Carlin (JASA 2007) that included 50 patients, I am no longer able to 
identify the 50 patients who were included.  Moral of the story:  Document this stuff while 
you're doing it, because you probably won't be able to reconstruct it later.  

The 13 columns in the dataset are:

-- Patient ID:  Subject ID number

-- Arch:  Upper (Maxilla) or lower (Mandible)

-- Tooth:  Tooth label using the World Dental Federation notation (http://en.wikipedia.org/
wiki/FDI_World_Dental_Federation_notation).  Each tooth's label has the form XY, where X 
indicates the quadrant and Y the tooth in the quadrant.  The quadrants are 

    1:  Upper right
    2:  Upper left
    3:  Lower left  (note!  They circle around the mouth)
    4:  Lower right

The teeth are numbered from the front/center toward the rear:

    1:  Central incisor
    2:  Lateral incisor
    3:  Canine
    4:  First premolar (aka first bicuspid)
    5:  Second premolar
    6:  First molar
    7:  Second molar
    8:  Third molar (aka wisdom tooth;  not included in this dataset)

Thus the tooth labelled 36 is the lower left first molar.

-- side:  Inner or outer side of the tooth.  Values are:  Lingual (tongue or inner side) 
and Buccal (cheek or outer side)

-- site:  Measurement location on the Arch, tooth, and site.  Values are Distal (farthest 
from the front/center of the mouth), Direct (in the middle of the tooth), and Mesial 
(closest to the front/center of the mouth).  

-- ToothSortInArch, SiteSortInToothNArch:  sort orders, as described above.  The file is 
sorted with mandible before maxilla and buccal before lingual, each of which is opposite to 
the smaller dataset.

-- al1:  Attachment loss measurement at that site (distance from the cemento-enamel 
junction to the bottom of the periodontal pocket)

-- pd1:  Pocket depth measurement at that site (distance from the free gingival margin to 
the bottom of the periodontal pocket).

-- age:  Age at the examination, in years

-- race:  Possible values:  Asian, Black, Caucasian, Hispanic

-- smoke:  Current smoker, Yes or No

-- sex:  Male or Female