The PartReg computer program (model) is designed
to perform a partial segmented linear
regression of one dependent variable
(Y, e.g. plant growth, crop yield) on one (X)
independent (explanatory) variable
(predictor), e.g. crop growth factors like depth
of water table and soil salinity with the aim to
find the water table and salt tolerance level,
also called the "no effect" range.
The segmentation is done by introducing a
breakpoint and finding the stretch (at the
beginning or at the end of the X-domain over
which the Y-X relation can be considered
horizontal, i.e. Y is not influenced by X. The
regression coefficient (i.e. the slope of the
line over the selected stretch) may not be
significantly different from zero in statistical
Two types of functions (3 and 4) are used.
In Type 3 the horizontal line is followed by a
sloping line whereas in type 4 the sloping line
is followed by a horizontal line.
The horizontal line is found by partial
regression over the head end (Type 3) or
the tail end (Type 4) of the X domain.
Difference with SegReg:
The PartReg program is similar to
SegReg. However, SegReg
minimizes the deviations of the calculated
values from the observed values over the entire
domain and comes with a broken line fitting all
data as well as possible. PartReg, on the other
hand considers only that part of the domain
over which the relation is horizontal. As a
result the stretch over which the line is
horizontal is usually longer than with SegReg,
but the fit of the data in the sloping segment
is not so good.
An illustration of the difference follows. For
more examples see
this article on
page 13 and following.
The salt tolerance (or response) of field
crops under farming
conditions in irrigated lands is analyzed in
tolerance paper for
cases in Egypt, India, and Pakistan.
reports & cases
reports & cases