[Gdpdm-devel] treatment tables in GDPDM

[Peter B. <pj...@co...>]

I was just noticing that the treatment tables and the trait tables in GDPDM 
have exactly the same structure.  There are a few more fields in the trait 
tables, but treatment data could be stored in the trait tables with no 
problem.  Also, in some cases, the distinction between treatments and 
traits is blurred.  A concrete example is some data that I helped upload a 
while ago.  It was data collected as part of an aluminum tolerance 
experiment.  Each taxon had two treatments (Al vs noAl) and root length was 
measured on three days.  The root growth measurement was hard on the 
plants, so it was basically a destructive measurement, so different plants 
were measured on each day.

The resulting data could be uploaded in a few different ways:
Aluminum and day as treatment, root length as trait.
Aluminum as treatment, day and root length as trait.
All three as traits.
Aluminum treatment as block in the plots table, day as either trait or 
treatment, root length as treatment.
day1_root_length, day2_root_length, day3_root_length as different traits

All those ways of loading the data would work.  However, to retrieve the 
data, I have to know something about how it was stored.  Either that or you 
always look in both the treatment and trait tables for whatever is 
there.  If you always have to check both tables, you might as well just 
store it all in one and make things simpler.

Other items that blur the distinction between treatment and trait:
In the above analysis you might want to treat day as a categorical variable 
or as a continuous variable ( a covariate).
In general, days to an event could be considered a treatment as in the 
above example, or as a trait as in days to silk.
Treatment and trait data can either be continuous or discrete.

The above argues pretty strongly that the treatment and trait tables should 
be merged, which simplifies both uploading and downloading.  For the 
ontology folks it also suggests that there should not be separate trait and 
environment ontologies.

Have a happy New Year.



Peter Bradbury
USDA-ARS
741 Rhodes Hall
Cornell University
Ithaca, NY  14853
607-255-5392