Journal Volume: 70      No.: 3     Year: 2016
S.No Title Abstract Download
1 Estimation of Parameters in Nonlinear Regression Models with Unequally Spaced Observations
Author: Trijya Singh      Pages: 189-195
In nonlinear regression models, the least squares estimates of parameters are obtained using non-linear optimization algorithms. These algorithms need good initial estimates as seed values to start iterations. Usually graphical methods or a combination of graphical and transformation methods is used for the purpose. But in some ill conditioned situations with these estimates as initial seed values, convergence may be slow or to a local minimum. It is also possible that convergence may not occur at all. Therefore we not only need a good set of estimates close to the true least squares estimates but we also need another set of estimates to ensure convergence to a global minima. For unequally spaced observations, there is no procedure in literature to find the initial estimates. In this paper, we have developed a procedure which provides good initial estimates for parameters of nonlinear regression models. The method could be applied to both equally as well as unequally spaced observations. We have applied the procedure to some published data sets to demonstrate that it procedures good initial estimates for optimization algorithms. Keywords: Non-linear Regression, Method of Least Squares, Asymptotic Regression, Growth Curves, Optimization Algorithms.
2 An Improved Two Stage Optional RRT Model
Author: Neeraj Tiwari and Prachi Mehta      Pages: 197-203
Randomized Response Technique (RRT) is an effective survey method for collecting data on sensitive issues such as drug uses, tax evasion and induced abortion etc., while trying to maintain respondent?s anonymity. Originally, RRT has been proposed by Warner (1965). Gupta (2001) and Sihm and Gupta (2014) considered two unknown parameters ? and ? in their model and estimated them taking two independent samples. In this paper, we have proposed an improved methodology for RRT, in which the sensitivity level (?) is considered to be known and the RR technique was applied only for those respondents who considered the particular question to be sensitive. This makes the procedure simpler and more efficient compared to the procedure of Gupta (2001) and Sihm and Gupta (2014). It has been theoretically established that the variance of the proposed estimator is less than the variance of the estimators suggested by Warner (1965) and Mangat and Singh (1990) under specific conditions. Some numerical examples have also been considered to demonstrate the utility of the proposed procedure over the existing RRT procedures. Keywords: Optional randomized response model, Randomization device, Sensitivity level, Scrambling variable.
3 Non-Linear Mixed Effect Models for Estimation of Growth Parameters in Goats
Author: Pankaj Das, A.K. Paul and Ranjit Kumar Paul      Pages: 205-210
Modelling growth of an animal is a complex process, because it requires describing longitudinal measurements with few parameters with biological interpretation. With longitudinal data, the variance of observations may increase with time (age), and repeated measurements of an individual over time are correlated. The non-independence of data violates a key assumption underlying many statistical procedures and has been ignored in most traditional non-linear fixed effect models. A solution to this problem is the use of non-linear mixed effect models (NLMM). A NLMM makes it possible to account for random covariates before testing for fixed effects and control autocorrelation in repeated measures. In this study, growth data of Goat has been used. Attempt has been made to develop the Von-bertalanffy mixed model. Logistic, Gompertz and Von-bertalanffy fixed and mixed models have also been explored for these data. Comparison of the models i.e. between fixed and mixed type of the same model and among different fixed and mixed models has been attempted. The goodness of fit statistics like i.e. Mean Square Error (MSE) and Root Mean Square Error (RMSE) of the fitted models has been computed. The parameters of the best fitted models along with their corresponding standard error are estimated. The performance of mixed effect models was found to be better than the fixed effect model. Specifically, under the category of mixed effect model, the Logistic model out performed over the other types that were considered in the study. Keywords: DM test, NLMM, Longitudinal data, Random covariates.
4 Genetic Variability of Growth Curve Parameters in Goats: Application of Bootstrap Techniques
Author: A.K. Paul, Ranjit Kumar Paul, N. Mohan Das Singh, S.D. Wahi and N. Okendro Singh      Pages: 211-218
Growth is an important phase in the life of animals which influences the different forms of production such as milk, meat etc. Since a series of weight-age data points are analytically difficult to interpret, it is desirable to study the growth of animals statistically. Inheritance of growth curves is critical for understanding evolutionary change and formulating efficient breeding plans. The genetic parameters are necessary to examine the potential usefulness of the growth parameters as selection criteria. In the present study, four models namely Logistic, Von berttallanffy, Gompertz and Weibull models are fitted to the body weight data of the goats of 142 animals. The Von bertalanffy model comes out to be the best model for the data under consideration. To this end, the statistical properties of the growth curve parameters are discussed by using bootstrap techniques and the distributions of these genetic parameters are found to be non-normal. The genetic correlation between the mature weight and maturity rate is found to be moderately negative correlated which indicates the selection of animals having higher maturity rate could lead to lighter mature weight. The heritability of mature weight is found to be highly heritable indicating the mature weight can be used for selection purposes. Keywords: Growth curve parameters, Heritability, Genetic correlation, Bootstrap techniques.
5 Calibration Approach based Estimation of Finite Population Total under Two Stage Sampling
Author: Kaustav Aditya, U.C. Sud and Hukum Chandra      Pages: 219-226
Auxiliary information is often used to improve the precision of estimators of finite population total. Calibration approach is widely used for making efficient use of auxiliary information in survey estimation. We proposed the regression type estimators of the population total using the calibration approach under the assumption that the population level auxiliary information is available at secondary stage unit level under two stage sampling design. The variance and the estimator of the variance of the proposed estimators were developed. We carried out limited simulation studies to demonstrate the empirical performance of proposed estimators. Our empirical results show that the proposed estimators outperforms the usual regression estimator under two stage sampling design in terms of the criteria of relative bias and relative root mean square error. Keywords: Auxiliary information, Calibration approach, Regression type estimator, Secondary stage unit, Two stage sampling.
6 Improved Estimation in Logistic Regression through Quadratic Bootstrap Approach: An Application in Agricultural Ergonomics
Author: Arpan Bhowmik, Ramasubramanian V., Anil Rai, Adarsh Kumar and Madan Gopal Kundu      Pages: 227-235
Keywords: Auxiliary information, Calibration approach, Regression type estimator, Secondary stage unit, Two stage sampling. based on Claeskens et al. (2003) for classification of data pertaining to the area of agricultural ergonomics have been discussed. Here, presence or absence of discomfort for the farm labourers in operating farm machineries has been considered as the dependent variable. presence or absence of discomfort for the farm labourers in operating farm machineries has been considered as the dependent variable. likelihood estimator and the quadratic bootstrap based estimator have been made based on real experimental situation in the field of agricultural ergonomics. The performance of quadratic bootstrap based estimator has been found to be better both in terms of length of the confidence interval of the parameter and classificatory ability of the model. Further, a bias corrected estimator based on quadratic bootstrap estimator following Claeskens et al. (2003) has also been obtained. A simulation study has been carried out which illustrates the improvement of bias corrected estimation over the usual maximum likelihood approach in terms of mean square error of the estimators and efficiency factor. Keywords: Bias correction, Classificatory power, Confidence interval, Ergonomics, Quadratic bootstrap, Simulation.
7 Weed Spatial Variability in Field Condition as Predicted by Kriging
Author: Yogita Gharde, K.K. Barman and P.K. Singh      Pages: 237-241
Keywords: Bias correction, Classificatory power, Confidence interval, Ergonomics, Quadratic bootstrap, Simulation. kharif 2012 to check the suitability of Randomized Block Design for weed control trials. Soil samples were collected from 33 grids (each of size 5 5 m2 season. Weed distribution maps were obtained through geo-statistical technique called kriging. These maps were prepared by using both weed count data in field as well as data obtained from weed seed bank study. Ludwigia parviflora accounted for approximately 50-60% of the total weed counts in the study season. However, total weed count data was used for modelling and kriging purposes. Result showed that the distribution of weeds is random in field situation and do not show any direction of gradient and thus violates the assumption of Randomized Block Design which is generally used for conducting weed control trials. Keywords: Geo-statistics, Kriging, Semivariogram, Spatial variability, Weed count.
8 Long Memory in Conditional Variance
Author: Ranjit Kumar Paul, Bishal Gurung, A.K. Paul and Sandipan Samanta      Pages: 243-254
Keywords: Geo-statistics, Kriging, Semivariogram, Spatial variability, Weed count. resulted strong evidence of long range dependence in the volatility processes for the series. Accordingly, FIGARCH model has been applied for forecasting the volatility of gram price. GARCH model and several extensions of GARCH models such as TARCH, EGARCH, Component GARCH and Asymmetric component GARCH have been applied for modelling and forecasting of return series. Evaluation of forecasting has been carried out separately in six moving windows by the help of mean squares prediction error (MSPE), mean absolute prediction error (MAPE) and relative mean absolute prediction error (RMAPE). The residuals of the fitted models were used for diagnostic checking. Diebold Mariano test was conducted for different pairs of models to test for the difference in predictive accuracy. It is found that FIGARCH model has better predictive accuracy as compared to all other models. It is also observed that component GARCH and asymmetric component GARCH models have better predictive accuracy than models. It is also observed that component GARCH and asymmetric component GARCH models have better predictive accuracy than models. It is also observed that component GARCH and asymmetric component GARCH models have better predictive accuracy than GARCH, TARCH and EGARCH models whereas there is no significant difference in the predictive accuracy of GARCH, TARCH and EGARCH models. The R software package has been used for data analysis. Keywords: Conditional heteroscedasticity, Gram price, Return series, Stationarity, Validation.
9 On Estimation of Population Mean under Systematic Sampling in the Presence of a Polynomial Trend
Author: Morteza Amini and Shahen Mohammad Faraj      Pages: 255-264
In this paper, we discuss the conditions under which a finite population might exhibit a polynomial trend of order k ? 1. The problem of estimation of a finite population mean in the presense of cubic trend is considered and a corrected estimator is obtained under a general systematic sampling scheme, which unifies the notations of linear, balanced and modified systematic sampling strategies. The performance of the estimator is evaluated using a super-population model approach. Assuming a cubic trend model the effect of a polynomial trend of higher order is evaluated by comparing the estimators developed under parabolic and cubic trend models. The real data example of grain production in Iran is considered to compare the performance of estimators. The numerical comparisons and real data analysis suggest that estimation of the population mean under the assumption that the population exhibits a cubic trend might result to better performance of the estimators rather than under those of linear or parabolic trends. Keywords: Empirical distribution function, Grain production, Relative efficiency, Weighted estimator.
10 Shrinkage Estimator in Dual Frame Survey Sampling
Author: Stephen A. Sedory, Sarjinder Singh and Maria del Mar Rueda      Pages: 265-275
In this paper, we have proposed a new shrinkage estimator of the population total for the dual frame survey sampling. The estimator is optimized for three optimizing parameters during the estimation process of the required population total. Thus it is likely that the proposed estimator will be more efficient than the Hartley (1962,1974) estimator listed in Lohr (2011) which makes use of only one optimization. At the end, a similar improvement of Fuller and Burmeister (1972) estimator is suggested. Keywords: Estimation of total, Dual frame surveys, Efficiency.
11 Development of Non-linear Models for Forecasting
Author: Anil Kumar and Sanjeev Panwar      Pages: 277-285
An attempt has been made to develop non-linear models for forecasting purpose. In the proposed models, more than one explanatory variables have been used. Two approaches are considered for building up the models. In the first approach, exact relation between the response variable and the explanatory variables are linearly added to get the final model. In the second approach, time variable of a nonlinear model is directly replaced by a linear relationship of explanatory variables. Proposed models are illustrated with a real life data of aphid population over time. It has been found that the model based on the first approach gives better forecast results. Keywords: Non-linear model, Forecast, Aphid, Mustard, Estimation, Linear regression.
12 A Web-based System on Herbicide Recommendation for Field Crops in India
Author: Chayna Jana, Shashi Dahiya , V.K. Mahajan , T.K. Das and N.M. Alam      Pages: 287-294
Weed control is a prerequisite for obtaining higher yield in agriculture. Amongst several weed control measures, chemical methods using herbicides have been proved to be the most effective and economical though time consuming. The potential effects of herbicides strongly depends on right selection relating to their toxic mode of action and method of application. Therefore appropriate herbicide should be recommended at correct dose and time of application to increase weed control efficiency. To provide a support, a web based application was designed and developed which will work as an online reference providing necessary information about various field crops, weeds, herbicide details, and appropriate herbicide recommendations across weeds and crops. It also provides scientific expertise to researchers, students, extension personnel and the end users nation wide towards the weed management in field crops using herbicides. Keywords: Web based system, Weed, Herbicide, Phytotoxicity.
13 Hindi Supplement
Author: ISAS      Pages: 6
14 Acknowledgement to Reviewers
Author: ISAS      Pages: 1
15 Summaries
Author: ISAS      Pages: 8