| タイトル | Evaluation of the procedure 1A component of the 1980 US/Canada wheat and barley exploratory experiment |
| 本文（外部サイト） | http://hdl.handle.net/2060/19820016678 |
| 著者（英） | Carnes, J. G.; Chapman, G. M. |
| 著者所属（英） | Lockheed Engineering and Management Services Co., Inc. |
| 発行日 | 1981-12-01 |
| 言語 | eng |
| 内容記述 | Several techniques which use clusters generated by a new clustering algorithm, CLASSY, are proposed as alternatives to random sampling to obtain greater precision in crop proportion estimation: (1) Proportional Allocation/relative count estimator (PA/RCE) uses proportional allocation of dots to clusters on the basis of cluster size and a relative count cluster level estimate; (2) Proportional Allocation/Bayes Estimator (PA/BE) uses proportional allocation of dots to clusters and a Bayesian cluster-level estimate; and (3) Bayes Sequential Allocation/Bayesian Estimator (BSA/BE) uses sequential allocation of dots to clusters and a Bayesian cluster level estimate. Clustering in an effective method in making proportion estimates. It is estimated that, to obtain the same precision with random sampling as obtained by the proportional sampling of 50 dots with an unbiased estimator, samples of 85 or 166 would need to be taken if dot sets with AI labels (integrated procedure) or ground truth labels, respectively were input. Dot reallocation provides dot sets that are unbiased. It is recommended that these proportion estimation techniques are maintained, particularly the PA/BE because it provides the greatest precision. |
| NASA分類 | EARTH RESOURCES AND REMOTE SENSING |
| レポートNO | 82N24554
NASA-CR-167566
E82-10276
JSC-17806
FC-L1-04219
NAS 1.26:167566
LEMSCO-16311 |
| 権利 | No Copyright |
