Statistical models for the assessment of yield loss due to weeds
By: Priyalakshmi M.
Contributor(s): Prabhakaran P V ( Guide).
Material type:
BookPublisher: Vellanikkara Department of Agricultural Statistics, College of Horticulture 2002DDC classification: 640 Online resources: Click here to access online Dissertation note: MSc Abstract: A study was undertaken to identify suitable functional models for
assessing the effect of weeds on the yields of three major crops of
Kerala Viz. Rice, Tapioca and Sesame and to estimate the loss in yield
in these crops caused -by the major weeds. The data required for the
study were gathered from the available records of A.I.C.R.P on weed
control . Multivariate techniques such as multiple linear regression
analysis, step wise regression analysis and principal component analysis
were used along with univariate techniques for the prediction of yield
and yield loss. The study undoubtedly revealed the importance of weed
in suppressing the potential yield of plants. The effect of weeds on
crops depended on the type of management , crop and season. Crop
loss estimates showed wide variation between seasons and locations.
The estimate of loss ranged from 5.3% to 68.4% in rice, 3l.4% to
46.3% in sesame and '12.8% to 40.6% in tapioca. The percentage of
avoidable loss' in different crops varied from 5.3% to 93.4%. Weed dry
matter (W.D.M.) was found to be the most important weed character in
, '
predicting crop yield and yield loss. Echinocloa was found to be one of
the major weeds causing considerable havoc to rice crop. In general
non linear mod~ls were more efficient than linear model in predicting
crop yield. The cauchy function, reciprocal hyperbola, second order
hyperbola and reciprocal straight line were adjudged to be the most
prormsmg univariate functional models in des£ribing the yield-weed
relation ship. Multivariate regression models were found to be more
powerful in predicting crop yield than univariate models. In most of the
cases the fitted statistical models described the proposed relation ship
with satisfactorily high degree of precision.
| Item type | Current location | Call number | Status | Date due | Barcode |
|---|---|---|---|---|---|
Theses
|
KAU Central Library, Thrissur Theses | 640 PRI/ST (Browse shelf) | Available | 171956 |
MSc
A study was undertaken to identify suitable functional models for
assessing the effect of weeds on the yields of three major crops of
Kerala Viz. Rice, Tapioca and Sesame and to estimate the loss in yield
in these crops caused -by the major weeds. The data required for the
study were gathered from the available records of A.I.C.R.P on weed
control . Multivariate techniques such as multiple linear regression
analysis, step wise regression analysis and principal component analysis
were used along with univariate techniques for the prediction of yield
and yield loss. The study undoubtedly revealed the importance of weed
in suppressing the potential yield of plants. The effect of weeds on
crops depended on the type of management , crop and season. Crop
loss estimates showed wide variation between seasons and locations.
The estimate of loss ranged from 5.3% to 68.4% in rice, 3l.4% to
46.3% in sesame and '12.8% to 40.6% in tapioca. The percentage of
avoidable loss' in different crops varied from 5.3% to 93.4%. Weed dry
matter (W.D.M.) was found to be the most important weed character in
, '
predicting crop yield and yield loss. Echinocloa was found to be one of
the major weeds causing considerable havoc to rice crop. In general
non linear mod~ls were more efficient than linear model in predicting
crop yield. The cauchy function, reciprocal hyperbola, second order
hyperbola and reciprocal straight line were adjudged to be the most
prormsmg univariate functional models in des£ribing the yield-weed
relation ship. Multivariate regression models were found to be more
powerful in predicting crop yield than univariate models. In most of the
cases the fitted statistical models described the proposed relation ship
with satisfactorily high degree of precision.
There are no comments for this item.