Improvement on Arima model forecasts for savings and credit co-operative societies in Wakiso District Uganda (2008 — 2012)
Loading...
Date
2013-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Kampala International University,school of Postgraduate studies and research
Abstract
There were five approaches to forecasting based on time series data namely exponential smoothing methods, single equation regression models, simultaneous equation regression model, ARIMA models and VAR models. Since SACCOs dealt with time series data the researcher selected the ARIMA modeling approach in the research. The research was carried out on SACCOs
in Wakiso District Uganda. The study specifically addressed the inadequacies of missing data using the Inverse Probability Weighing method and the Round Table Imputation method that was invented by the researcher, the software package designed to handle data for variables with equal and unequal number of observations, run an automated statiEtical tables system
that enabled the setting of critical regions for the various statistical tests, carried out the tests for normality, linearity, eteroscedasticity and
stationarity in an orderly manner according to the Classical Linear Regression
Model and gave conclusions for those tests. The automated system provided
seven ARIMA models from which the best fit was selected. The findings of
the study were: Round Table Imputation method was found easy to apply
and its results were consistent with the expectations, the Inverse Probability
Weighting method was also found to be effective, SACCOs were able use the
invented EOI to check their effectiveness at a glance and identify the areas of
inadequacies, automated statistical tables were used which was convenient
to the users to carry out various tests with ease and the availability of
conclusions from the tests made the statistical package supportive in making
conclusions. In conclusion, missing values in the data whether missing at
random or not should be addressed before use for forecasting, unequal
observations for the different variables were also considered no longer a
hindrance to data analysis, automated statistical tables were useful and
eliminated the omission and commission of picking the critical values
manually and statistical tests that were carried out before the data was used
for forecasting led to better results. It was recommended that when
encountered with missing data the reason for missingness should be
established and if it was found that the missing was at random remedial
measures should be used to fill in the missing data but if the missingness was
found not to be at random the missingness component should be included in
the forecasting model, users of data that was of time series nature should
use statistical computing package such as Herbo Arima which was an
improvement of ARIMA data modeling system, the automated system was
recommended for those users who may wish to do the forecasting without
much ado as to the nature of the data they handled, statistical computing
packages should have facility for users to access automated statistical tables
to minimize on omission and commission errors and SACCOs should be able
to use the Effective Operating Index provided in the computing package to
monitor their effectiveness in the market.
Description
Thesis Presented to the School of Postgraduate Studies and Research Kampala International University. In Partial Fulfillment of the Requirements for the Degree Master of Science in Statistics
Keywords
Improvement on Arima model forecasts, Savings and credit co-operative societies, Wakiso District, Uganda