06/16/2000: TDT Documentation in MS Word
Please read Transmission_disequilibrium_tests.doc in MS Word
format. It was created by Dr. Chris Amos

9/29/1998: Updates
There is no need for alleles to be down coded. compute7.sas has been
modified today to take real allele symbles (numeric please!) into
TDT calculation. Final output will show this change. Allele symbols
may not be sequential numbers any more.

1. Introduction on TDT:

This is a description on one-marker TDT calculation.  Model programs 
are tdt6.sas and compute7.sas as of today (June 12, 1996). Simulation 
part has a separate explanation which will be implemented soon. In 
this version, single parent structures are dealt with, an error 
checking scheme is introduced. To run this program successfully, we
recommend you have SAGE available, or a program produces nuclear
family structures like SAGE/FSP does.

a. Single parent stucture:
In a single parent case, suppose the father's type is 12, mother 00
(missing) and a kid has a type of 12.  Then first, 1 is assumed
transmitted, 2 not transmitted, with a weight of 0.5; second, 2 is
assumed transmitted and 1 is not transmitted, also carrying a weight
0.5. For other cases where definitive transmission status is 
obtainable, mark the transmission pattern with such shuffling and 
assign a weight of 1. For example, if father has 12, mother 00
(missing), kid has 13, then 1 is definitly considered transmitted, 3
not transmitted, with a weight of 1, and that's the end of the
story.

b. The error checker:
It picks up all the "left-overs" after a genotype matching is exhausted,
meaning, a kid's given allele has failed to find a match from any four
(4) of the parental allele. In this case, the user needs to go back to
the data and fix the error before any meaningful calculation can go
forward.  For the time being, SAS will stop operating and posts an
error message in both its output files (.log and .lst).

c. Data:
For data preparation, user should prepare the data into a standard
pedigree structure. In a typical 2-dimensional table, there could be
some columns (variables) as so listed below:

famid       (family id)
id          (person id within each family)
faid        (father's id)
moid        (mother's id)
sex         (individual's gender)
aff         (affection code, 0=missing, 1=non-affected, 2=affected)
m11         (marker 1 allele 1)
m12         (marker 1 allele 2)
m21         (marker 2 allele 1)
m22         (marker 2 allele 2)

Each line will in turn represent a person. For now, TDT only works with
nuclear family structure. If the original data is organized in extended
family fashion (grandparents+parents+siblings), then the data should be
preprocessed into nuclear family structure first.  Currently, SAGE
provides a program FSP to cut large families into nuclear pieces.

d. The program:
The core file is compute7.sas. There has been at least 6 major updates
since the beginning of the project. It is in turn called by a master 
program tdt6.sas. The latter is nothing but a calling engine. More
detailed words can be found from within the programs themselves.

2. Example:

We figured that the best way to use this program is to follow an
example, and observe every step along the way. We borrowed some data
from England to illustrate this. The original data shipped to us in
Microsoft Execl format and have been converted into a standard ASCII
file called raw.dat. Let's start with raw.dat.

a. run sas t.sas
   this would produce a sas dataset file called t.ssd01. It also 
   produces a by-product called marker.names, for an executable
   program to pick up. Executable program t in trun in created from a
   FORTRAN program t.f.

   The most important output file is an ASCII file fsp.dat. This can
   be linked to fort.11 for fsp to run. Remember to change -9 into
   two spaces using your favorite editor.

   NOTE: Modify marker.names immediately to delete lines does not
   contain allele names. A valid marker.names file should look like
   this:

   a1
   a2
   c1
   c2
   dr1
   dr2

   The above lines suggests that you have three (3) markers, a, c and
   dr, each is represented by two alleles (next to each other).

b. Compose your fsp.par file and link it to fort.1, then run fsp to 
   get fort.23, which has nuclear family structure

   NOTE: Always check to see if fort.21 smokes. Multiple-mating is
   O.K. though.

c. run read23 to convert fort.23 into tdt.dat, an ASCII file retains
   the nuclear family structure but add a sequential family id into
   the file. 

d. run sas nuc.sas to convert tdt.dat file into a sas dataset file
   called t.ssd01.

e. run t, executable from t.f
   this would produce a file called reorder.sas, to be included in
   t1.sas. This program also prepares variable labels for marker
   alleles.

f. run sas t1.sas
   This would reorder allele sequencies to a bounch of sequencial
   numbers, so that tdt can process later. A permanent sas data file
   called t1.ssd01 is produced afterwards. By product t1.lst is
   necessary to show the mapping schems. 

g. run tdt1.sas, tdt6.sas and tdt8.sas, using t1.ssd01 as input file
   and conduct different tdt tests

h. A detailed explanation on tdt8.sas and sim8.sas will be provided
   later. For the time being, contact Dr. Amos at
   camos@request.mdacc.tmc.edu.


README file prepared 10/01/1997
Last Updated: 11/06/1997