Statistics

Pandey (1977) considered the exponential distribution which is generally used in life testing experiment. Epstein & Sobel (1953) and Bhattacharya & Srivastava (1974) considered the censoring procedure in life testing problem. Pandey and Malik (1994) considered the improved estimator for ?2 in exponential distribution. In this paper proposed an estimator for ?2 in case of exponential distribution and studied its properties under Linex loss function.

Many researchers have carried out research on students’ academic performances in the University system, but there is dearth of information about models which focus on the relationship between students pre and post admission performances. This paper was therefore designed to model relationship between students’ pre and post admission performances. Information on pre admission performances (Olevel, Jamb and Post-Jamb results) and post-admission performances (100L – 400L results) of students’ in Statistics Department, Federal University of Agriculture, Abeokuta, Ogun State, Nigeria was collected from students’ file. Correlation matrix and Canonical Correlation analysis were used to know the degree of relationship that exists between the pre and post admission performances. The principal component analysis was employed to reduce the multidimensional data. Scree plot was used to determine the spread of the trend of the components and bi plot was used to determine the degree of closeness of the students’ pre and post admission performances. There is no relationship between pre and post admission performances. Also, no strong relationship among pre admission performances, while the relationship among post admission performances is very high. Post admission performances are highly related to students’ CGPA. The proportion of variance accounted for by the first, second and third principal Components are 50.7542%, 16.5712% and 15.6224% respectively with cumulative proportion of 82.95%. The first, second and third components are chosen. The seven components were reduced to three. Post admission performances are closely related and stand as the determinant of students’ class of degree.

The paper considered the construction of a family of BIBDs using potential Lotto Designs (LDs) earlier derived from qualifying parent BIBDs. We utilized Li’s condition, to determine the qualification of a parent BIBD as LD constrained on and then considered the case since is the smallest number of tickets that can guarantee a win in a lottery. The BIBD was used as the parent BIBD for the procedure. This BIBD yielded three potential LDs each of which was completely generated using a Microsoft Office Access database computer program and their properties were studied. The three LDs, after their complete generation, yielded the and BIBDs. These BIBDs follow the generalization where are the parameters of the BIBD. A MATLAB program was used to generate a family of the BIBDs for with these new set of parameters. All the BIBDs in this family are unreduced designs.

In this paper, we establish some expectation identities satisfied by single and product moments of Generalized Order Statistics from Left Truncated Logistic Distribution. These identities are independent of left truncation point and therefore also applicable to Logistic as well as for half Logistic distributions studied in Balakrishnan (1985) and Saran and Pandey (2012) . A particular case of these results verify the corresponding results of Saran and Pandey (2004).

This paper aim to discuss the preliminary study of modelling algorithm setting of Stochastic Boosted Regression Trees (BRT) using R statistical software. The ground level ozone (O3) concentrations from a Malaysia’s air quality monitoring station was used as a case study. The new approach; BRT developed by Friedman (2001), Friedman (2002) and adopted by Yahaya (2013) in air pollution studies. BRT is differ substantially from traditional regression-based approaches. BRT are constructed of multiple regressions models which finally produce a single ‘best iteration’ model to optimize prediction performance. Sensitivity testing of model been carried out to determine the best parameters’ setting which are number of trees (nt) (1000-10000), learning rates (lr) (0.005-0.5), and interaction depth (tc) (1-10) to suits the ozone data. Results indicated that the BRT analysis algorithm best modeled with the best combination of parameters lr=0.001, tc=5 and nt=5301 that achieves minimum predictive error (minimum error for predictions). The algorithm model would crucially best provided clear benefit for air pollution study and their ability to model complex variable interactions and non-linear effects.

Some construction methods of 3-resolvable regular group divisible designs based on incidence matrix of known affine resolvable balanced incomplete block designs and 3-resolvable group divisible type partially balanced incomplete block designs with unequal block sizes within each resolution set based on incidence matrix of known symmetric balanced incomplete block designs are proposed with illustrations.

This paper deals with a cold standby redundant integrated h/w and s/w systems with priority, preventive maintenance, maximum operation time and maximum repair time. The system fails completely directly from normal mode. A single repair facility is provided who visits the system immediately. After, a maximum operation time system undergoes for preventive maintenance directly from normal mode. If repair of the h/w is not possible by the server up to a pre-specific time, it is replaced by new one with some replacement time. However, only up-gradation of the software by new one is made whenever s/w fails to execute the desired function properly. In model 1st, no priority is given whereas; in 2nd model, priority is given to s/w up-gradation over h/w repair. The failure time distribution of the h/w and s/w follows negative exponential while the distributions of preventive maintenance, repair, up-gradation and replacement time are taken as arbitrary with different probability density functions. The comparative study of mean time to system failure (MTSF) and profit function of both models have been made graphically with respect to preventive maintenance rate for fixed values of other parameters.

In this paper, Bayesian estimators of the population proportion of a sensitive attribute were developed when real life data were gathered through the administration of survey questionnaires on an induced abortion on 300 matured women in some selected hospitals in the metropolis. Using both the Kumaraswamy (KUMA) and the Generalised (GLS) beta distributions as alternative beta priors, efficiency of the proposed Bayesian estimators was established for a wide interval of the values of the population proportion. We observed that for small, medium as well as large sample sizes, the proposed Bayesian estimators were better than the conventional Bayesian estimator proposed by Hussain and Shabbir [10] when a simple beta prior was used.

A design of experiment is a plan to collect measurement or observation according to a pre arrange plan in such a way as to provide the basic for valid inference. This work was carried out to examine the research station effect on the yield of Cowpea varieties. The station are four locations in Nigeria (Kaduna, Shika, Mokwa and Kano). Eight different varieties of Cowpea were considered (Tg 1910-8F, Tg 1844 – 1E, Tg1019 – 2E, Tg1904 – 6F, Tg1910 – 2F, Tg1448 – 2E, Tg1908 – 1F, and Tg1740 – 2F). The data are secondary data, collected from International Institute of Tropical Agriculture (IITA) Ibadan, Oyo State. The result showed that research locations has no significant effect on the yields of cowpea varieties. The use of Randomized Complete Block Design (RCBD) design in Kaduna station, Shika station, Mokwa station and Kano station had 27.2%, 109.9%, 63.04% and 53.7% gain in experimental precision respectively.

This study examined and compared six basic time series forecasting models (Exponential model, Double Exponential model, Holt-Winter models, Time Series linear regression model, the ad-hoc Bootstrapping model and the Self Adjusting model) with application to twenty-four Months Nigeria’s CPI inflation sample data, from January 2009 to December 2010 inflation data. With the aids of five different standard forecasting accuracy measures (MSE, MAE, RMSE, SSE, and MAPE), results from the out-of-sample forecasts shows that the double exponential model with a smoothening constant of 0.68 is the best forecasting model for the Nigeria inflation rate data among the other ad-hoc model considered.