International Journal of Probability and Statistics 2016, 5(1): 10-17
DOI: 10.5923/j.ijps.20160501.02
Performance Assessment of Penalized Variable Selection
Methods Using Crop Yield Data from the Three Northern
Regions of Ghana
Smart A. Sarpong
1,*
, N. N. N. Nsowah-Nuamah
2
, Richard K. Avuglah
3
, Seungyoung Oh
4
, Youngjo Lee
4
1
Institute of Research, Innovations and Development - IRID, Kumasi Polytechnic, Ghana
2
Kumasi Polytechnic, Ghana
3
Department of Mathematics, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
4
Department of Statistics, Seoul National University, South Korea
Abstract Time and money can be saved by measuring only relevant predictors. Measuring relevant predictors also
ensures a noise free estimation of parameters and preserves some degrees of freedom for a given predictive model. By
comparing the performance of SCAD, LASSO and H-Likelihood, this study seeks to select among access to credit, training,
study tour, demonstrative practicals, networking events, post-harvest equipments, size of plot cultivated and number of
farmers; variables (fixed and interaction terms) that significantly influences crop yield in the three Northern regions of Ghana.
Our simulation as well as real life results gives evidence to the fact that H-Likelihood method of penalized variable selection
performs best followed by SCAD, with LASSO coming last. It does both selection of significant variables and estimation of
their coefficients simultaneously with the least penalize cross-validated errors compared to the SCAD and the LASSO. The
study therefore recommends that deliberate efforts be put into strengthening the Agricultural support systems as a form of
strategy for increasing crop production in Northern Ghana.
Keywords Penalized, SCAD, LASSO, H-likelihood, PCVE, Variable Selection, Crop yield
1. Introduction
The rate of food production in many parts of sub-Saharan
Africa has not kept pace with the rate of population growth.
Whereas the estimates of population growth rate increase at
about 3 per cent annually, that of food production increases
by only 2 per cent (Rosegrant et al., 2001). The sub-region's
per capita deficit in grains and cereals according to
Rosegrant et al., (2001) is one of the highest in the world.
Way back in 1967, the sub-region's cereal imports was 1.5
million tons. However, just within thirty years down the
way, this figure increased to 12 million tons in 1997, and
projections have it that the sub-region will require about 27
million tons of cereal imports to satisfy demand by 2020
(Rosegrant et al., 2001). In the long run, importation may
not be economically feasible to ameliorate food shortages.
Thus, there is a need to increase domestic production to
guarantee food security.
Ghana is still recognized worldwide as an
agriculture-based economy. Agriculture has been the anchor
* Corresponding author:
smartsarpong2015@gmail.com (Smart A. Sarpong)
Published online at http://journal.sapub.org/ijps
Copyright © 2016 Scientific & Academic Publishing. All Rights Reserved
of Ghana’s economy throughout post-independence history
(McKay and Aryeetey, 2004). While policy and political
instability had induced the fall of per capita GDP growth
until 1980s, the agricultural sector had been less affected due
to less interventions by the government compared to the
non-agricultural sector and the fact that its growth is mainly
led by smallholder farmers for subsistence purposes. Beyond
1992 when Ghana gained the forth republican political
stability, there has been a rapid growth in the nonagricultural
sectors; expanding by an average rate of 5.5 percent annually,
compared to 5.2 percent for the entire economy (Bogetic
et al., 2007).
The analysis presented in this paper suggests that a system
of support services; Access to credit facility, Training, Study
tour, Demonstrative practical, Networking events and
Post-harvest Equipments, plays an important role in
determining crop yields even though their individual and
interaction effects on yield is not uniform across farmer base
organizations. This research focuses primarily on the
production of Maize and Soy beans in northern region of
Ghana where there exists considerable farming activity.
Maize and Soy beans are the very much cultivated in these
parts of the country due to their vegetation which supports
the growth of grains and cereals. Beyond the numbers and
descriptive statistics on yield of such crops, this study tries to