Analysis of Data from Crop Protection Experiments Using Generalized Linear Model: the Case of Parthenium
No Thumbnail Available
Date
2008-07
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Abeba university
Abstract
Among many weeds that cause crop loss, parthenium was found to be the most terrible one
according to some exploratory studies. The problem of parthenium is not only that it cause very
sever crop loss, but also it cause health problems to human and animal beings. Control of
Parthenium by Farmers ,cultural and labour intensive, caused farmers to suffer from skin allergy,
itching, fever, and asthma.
This study tried to popularize different Generalized linear models for modeling agricultural data
which is used for describing the data sufficiently well and then identify the natural relationship
between different variables for further analayis as well as applications. Generalized linear models
(GLMs) are used to do regression modeling for non-normal data with a minimum of extra
complication compared with normal linear regression. One of the available programs that is
important in current statistical practice is the GLM procedure in the SAS software package.
The study is based on the result of a parthenium and other species count data, secondary
data,obtained from Ethiopian instisute of Agricultre research. Descriptive statistics supported by
graphical presentations have been discussed to show the dominance of parthenium on other species
per plot area. Furthermore, to evaluate the probability of a plot or a quadrant to be free of
parthenium, models form GLM family are applied to the data using SAS software.
Based on the parameter estimates, fitted models were formulated, parameters are interpreted and
comparison of fitted models conducted. In this model fitting process, an attempt was made to
alleviate a confusion of which model to which data. The logit and probit model fitting gives similar
results for the same data as expected and the choice of one model cannot be made based on AIC,
because the AIC for both models is the same.
The poisson regression model fit is found to be inadequate for two different variables, as its
Deviance value is far from one. The Negative Binomial Model gives a better fit and its Deviance
shows the model is adequate for the same data used for poisson regression. The multinomial logit
model for parthenium infestation in five categories as dependent variable and the sum of all other
species gives a better result, as infestation level increase i.e as the severity of parthenium infestation
increase, the number of the sum of other species gets low which in turn means that the probability of
getting other species gets very low
Description
Keywords
the Case of Parthenium