Loading...
Theses
Browse
2 results
Search Results
ThesisItem Open Access Resampling Techniques For Evaluating G × E Interaction In Oilseed Crops(Chaudhary Charan Singh Haryana Agricultural University hisar, 2022-09) Deepankar; Hooda, B. KIn multi-environment trials (METs), a set of genotypes is grown simultaneously in different set of environments. The major objective of METs is identification of genotypes which consistently perform across a wider range of environments. To assess the stability of genotype, in literature there exists various parametric and non-parametric measures. But researcher faces conundrum of choosing appropriate stability measure before moving to main objective of MET. To ease researcher in this dilemma, we developed majority approaches where the results of various parametric and nonparametric stability measure were combined. Under majority approaches, we evaluated i) rank sum of parametric and non-parametric stability measures, ii) modal approach, iii) A new weighted-normalized index and iv) a composite measure using TOPSIS algorithm. A statistical distribution is a mathematical function that describes how the results of an experimental trial are likely to occur at random. The stability measures are a complex function of observed values therefore it is difficult to develop theoretical framework to predict their sampling distribution. Hence bootstrap technique has been used to determine their sampling distributions of stability measures. In METs, for studying GEI, additive main effects and multiplicative interactions (AMMI) and genotype and genotype x environment interaction (GGE) models are frequently used by researchers. In both models after removing the additive effect, singular value decomposition is used to partition genotype x environment interaction into ordered sum of multiplicative terms. Researchers usually retain first two multiplicative terms for biplot analysis without giving much thought in checking the significance of multiplicative terms for retention. The resampling techniques such as bootstrap and cross-validation have been used to test the significance of the multiplicative terms by approximating p value for each multiplicative term. Only those multiplicative terms have been retained in model which are found to be significant i.e., p value < 0.05 or 0.01.ThesisItem Open Access Weather based rice yield prediction models for Karnal District of Haryana(CCSHAU, 2017) Deepankar; Aneja, D.R.The study carried out for Karnal district of Haryana was based on historical data for a period of 35 years (1980-81 to 2015-16). Rice yield data were collected from Statistical Abstract of Haryana and daily weather data were obtained from Central Soil Salinity Research Insitute (CSSRI), Karnal. Stepwise multiple regression technique has been applied for period 1981-82 to 2012-13 with yield as dependent variable and weather indices (artificial variables generated from weekly & fortnightly weather values) as independent variables. Another three years data (2013-14 to 2015-16) have been used for the validation of the models. The models based on maximum temperature (22 & 24 weeks) and no. of rainy days (22 & 24 weeks) are comparable with each other on the basis of adjusted R2, therefore on the basis of root mean square error the model based on no. of rainy days (24 weeks) having lowest RMSE (219.14) is chosen among all models based on individual weather variables. The actual forecasts using model based on no. of rainy days (24 weeks) for 2013-14 to 2015-16 years were 3207.11 kg/ha, 3318.09 kg/ha and 3322.61 kg/ha, respectively. The model based on joint effect of maximum temperature and relative humidity morning (22 weeks) forecasts rice yield were very close to the actual yields (per cent relative deviation ranging from 1.15 % to 7.6 %. The actual forecasts using maximum temperature and relative humidity morning for 2013-14 to 2015-16 years were, 3244 kg/ha, 3168.03 kg/ha and 3215.34 kg/ha, respectively.
