Trades in stock market anywhere in the world is faced with intense volatility due to stocks prices instability in real time that is mostly driven by information and other market dynamics. This research examines two volatility models with two different error distributions innovations in modelling and forecasting the continuous compounded return series (CCRS) of Nigeria All Share Index (NGX ASI) spot prices spanning the period of January 30, 2012 to June 30, 2021. The Generalized Autoregressive Conditional Heteroscedastic (GARCH) and Asymmetric Power Autoregressive Conditional Heteroscedastic ARCH (APARCH) volatility models under Student-t Distribution (StD) and Generalized Error Distribution (GED) error innovations are utilized. The best-fitted model is determined using Akaike’s Information Criterion (AIC) while Mean Square Error (MSE) is used to evaluate forecasts performance of the fitted volatility models. The results from the analysis showed that amongst competing models, APARCH (1,1)-GED was selected to be the best fitted volatility model with better forecasting power for the CCRS-NGX-ASI spot prices. This is because it produces the smallest AIC and MSE values

Abstract

The study compares two statistical methods: Discriminantt analysis with Logistic regression model inpredicting birth weight of an expectant mother. Normal birth weight and Low birth weight. 240 cases of (infants) was observed with the following measurements considered maternal height (x1), maternal weight (x2), maternal age (x3), baby’s weight (x4), baby's sex (x5), gestational age (x6) and parity (x7) of an expectant mother, Discriminant Analysis classified the Normal birth weight correctly (64.6%) while it recorded (64.7%) success rate in classifying the Low birth weight. In the case of the Logistic regression, it recorded (76.8%) and (52.9%) success rate in classifying the Normal birth weight and Low birth weight respectively. The overall predictive performance of the two models was high with the Logistic regression having the highest value (65.8%) Among the seven characteristics examined, Maternal height, maternal age, Baby’s weight, sex, gestational age and parity were not significant variables for identifying Birth weight by both methods while Mothers weight is important identifying variable for both except Mothers age which was significant in the Discriminant analysis. The study shows that both techniques estimated almost the same statistical significant coefficient and that the overall classification rate for both was good while either can be helpful in selection of birth weight however, given the failure rate to meet the underlying assumptions of Discriminant Analysis, Logistic Regression is preferable.