eduzhai > Physical Sciences > Earth Sciences >

Calibration of a Confidence Interval for a Classification Accuracy

  • carsar
  • (0) Download
  • 20210408
  • Save

... pages left unread,continue reading

Document pages: 23 pages

Abstract: Coverage of nominal 95 confidence intervals of aproportion estimated from a sample obtained under a complex survey design, or aproportion estimated from a ratio of two random variables, can departsignificantly from its target. Effective calibration methods exist forintervals for a proportion derived from a single binary study variable, but notfor estimates of thematic classification accuracy. To promote a calibration ofconfidence intervals within the context of land-cover mapping, this study firstillustrates a common problem of under and over-coverage with standardconfidence intervals, and then proposes a simple and fast calibration that moreoften than not will improve coverage. The demonstration is with simulatedsampling from a classified map with four classes, and a reference class knownfor every unit in a population of 160,000 units arranged in a square array. Thesimulations include four common probability sampling designs for accuracyassessment, and three sample sizes. Statistically significant over- andunder-coverage was present in estimates of user’s (UA) and producer’s accuracy(PA) as well as in estimates of class area proportion. A calibration with Bayesintervals for UA and PA was most efficient with smaller sample sizes and twocluster sampling designs.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×