Thumbnail Image

Theses

Browse

20234
Reset filters
Subject: Statistics × End date: 2023 × Start date: 2023 × Type: Thesis × Thesis Type: Ph.D ×
Reset filters

Search Results

Now showing 1 - 4 of 4
  • Thumbnail Image
    ThesisItemOpen Access
    Volatility forecast models of prices and arrivals of tomato in APMC markets of Haryana
    (CCSHAU, Hisar, 2023-01) Pushpa; Joginder
    The majority of agricultural time series data are nonlinear, nonstationary and leptokurtic in nature. Thus, one of the most difficult areas of time series forecasting is agricultural price forecasting. Accurate forecasting assists both farmers and policymakers in making good decisions. According to the literature, each of the forecasting models has its own set of limitations. In the current study, forecasting performance of SARIMA, GARCH, ANN, Hybrid (SARIMA-GARCH and SARIMA-ANN) and multivariate time series (VAR and VARMA) models has been compared for monthly prices and arrivals of tomato in selected markets of Haryana. The purpose of the study is to give short term forecast of prices and arrivals of tomato with various forecast horizons such as one, three, six, nine and twelve months. Based on empirical results of the study, it is found that ANN models outperformed the others models for all horizon except one month ahead based on performance measures like MAPE and SEP. It is observed that Hybrid (SARIAM-ANN) models do not enhance the forecasting performance. The hybrid (SARIMA-GARCH) model outperforms the individual SARIMA and GARCH models in forecasting the prices and arrivals of tomato. It can be seen that the residuals obtained from linear SARIMA models contain the appropriate ARCH effect. The results of multivariate time series reveal that VARMA model outperforms the VAR model based on minimum values of forecasting performance measures such as MAPE and SEP.
  • Thumbnail Image
    ThesisItemOpen Access
    Modeling of tuberculosis through structural equations and bayesian approach
    (CCSHAU, Hisar, 2023-01) Rohit Kundu; Sheoran, O.P.
    This study analyzed data on tuberculosis in India to identify latent variables and understand relationships between variables. A structural equation model (SEM) was used, but the initial model did not converge to a satisfactory solution. The model was revised and modified until it converged to an optimal solution with acceptable fit statistics. The Markov Chain Monte Carlo (MCMC) method was also used to identify change points in the number of notified cases of tuberculosis. The results showed an increase in TB cases in the 2000s, followed by two change points after 2010 when the government prioritized controlling the disease. However, the number of cases has continued to increase in recent years. The MCMC method and Gibbs sampler were found to be useful for analyzing epidemiological data with change points. The study also analyzed the prevalence of TB in India using data from the National Family Health Surveys from 2005-2006 and 2015-2016. The results showed that overall, the prevalence of TB did not significantly change between the two surveys. However, the gender gap in TB prevalence (difference in prevalence between males and females) did show a statistically significant decrease, mainly observed in rural areas and found to vary by religion and social group. The rural-urban gap in TB prevalence was most prominent among certain groups, including Muslims, individuals belonging to other religious groups, Scheduled Tribes, and those in the poorest wealth quintile. It is suggested that the decreasing trend in the gender gap may be due to an improvement in the socio-economic status of women and increased detection and reporting of TB cases among women.
  • Thumbnail Image
    ThesisItemOpen Access
    A study on impact of climate change and statistical models for pre-harvest forecast of wheat-yield in Haryana
    (CCSHAU, Hisar, 2023-05) Chetna; Monika Devi
    This study aimed to improve the predictability of wheat yield in four districts of Haryana state using advanced statistical techniques. The best models for predicting weather variables were identified, and analyzed the impact of weather variables on crop yield during different growth stages. It was found that weather variables had varying effects on crop yield during different growth stages and across different districts. The observed positive effects of temperature on crop yield during the reproductive stages could be attributed to increased photosynthesis and growth rate of the crop, while the negative effects of temperature during the germination, milking, and harvesting stages could be due to increased plant stress and water loss. The study also found that the negative effects of rainfall on crop yield during certain growth stages could be attributed to waterlogging and soil compaction, while the positive effects of rainfall during certain growth stages could be due to increased soil moisture availability. The study developed models with high R2 values and low error values for predicting wheat yield in all four districts. Pre-harvest forecast models were developed to predict wheat yield before harvest in selected districts of Haryana, using discriminant function analysis and weekly meteorological variables. The models achieved high accuracy in correctly classifying the grouped cases in all districts, with varying effects of predictor variables and autocorrelation. The evaluation of various models for yield forecasting in different districts of Haryana State has yielded impressive results. Principal Component Analysis (PCA) was also utilized to investigate the impact of weather variables on the weather indices in various districts of Haryana State. The models showed a good fit with observed data and high accuracy in predicting yield, with different levels of complexity and performance depending on the district and the model used.
  • Thumbnail Image
    ThesisItemOpen Access
    Bayesian Estimation for Some Lifetime Models under Different Loss Functions
    (Chaudhary Charan Singh Haryana Agricultural University hisar, 2023-01) Pavitra Kumari; Vinay Kumar
    The life testing experiments are carried out to obtain the lifetime data on patients for survival analysis and to study the reliability of electrical, electronic and mechanical systems, information theory, artificial intelligence, etc. This thesis deals with the classical and Bayesian estimation methods for the generalized of lifetime models. We consider four distinct loss functions, namely, square error loss function, entropy loss function, precautionary loss function, Linex loss function and type II censoring in this thesis. Type II censoring has the significant advantage that you know in advance how many failure times your test will yield. Generalizations of univariate lifetime distributions are often of interest to serve for real life phenomena. These generalized lifetime distributions are very useful in many fields such as medicine, physics, engineering and biology. We consider three distinct lifetime models, namely, Lomax, Rayleigh Lomax and IPBH lifetime model and developed statistical inferences for the associated model parameters and reliability characteristics from both the classical and Bayesian estimation perspectives in Chapter 4. Lomax distribution is one of the well-known univariate distributions that is considered as an alternative to the exponential, gamma and Weibull distributions for heavy tailed data. In this thesis, we introduce a generalization of the Lomax distribution called Rayleigh Lomax (RL) distribution. This distribution provides great fit in modelling wide range of real data sets. It is a very flexible distribution that is related to some of the useful univariate distributions such as exponential, Weibull and Rayleigh distributions. Moreover, this distribution can also be transformed to a lifetime distribution which is applicable in many situations. For example, we obtain the inverse estimation and confidence intervals. In present study apply AIC and BIC to detect the changes in parameters of the RL distribution. The performance of these approaches is studied through simulations and applications to real data sets. The statistical software R is used for computation throughout the thesis. Finally, a complete list of references and other literature surveys are given at the end of the thesis as a bibliography.