The presence and characterization of measurement error in forestry Public Deposited

http://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/n583xx00g

Descriptions

Attribute NameValues
Creator
Abstract or Summary
  • Measurement error (ME) is a component of any study involving the use of actual measurements, but is often not recognized or is ignored. The consequences of MEs on estimates of tree and stand attributes and the parameters and predictions of forest models can be varied and severe, including the presence of bias and increased variance. While correction methods do exist for countering the effects of MEs, these methods require knowledge of the distribution of the errors. A new method for directly modeling error distributions called the two-stage error distribution (TSED) method is presented. This method is compared with traditional methods for error modeling through examples using diameter (D) and height (H) MEs. Comparisons are done based on a measure of dissimilarity between their fitted error distribution surfaces and the empirical error surface. Results indicate that the TSED method produces a much more accurate characterization of the ME distributions than traditional methods when a high percentage of errors are zero. When few measurements are exactly correct, the TSED method works as well as the most accurate form of the traditional method. The TSED method also performs better at characterizing distributions with asymmetric tails. It is therefore more adaptable than traditional methods and should be used for future error modeling. Variables included in forest models are often not simple variables such as D and H, but rather transformations of these variables. Applying correction techniques to account for MEs in these transformed variables requires distributions for the MEs of these variables. This information is often not available and may be time-consuming and expensive to collect and model. Indirect derivation methods are examined for obtaining the error distributions of transformed variables when the error distributions for their component variables are known. Statistical error distributions for D2 and D2H are first indirectly derived using known error distributions for D and H, and then directly estimated using the TSED method and under the assumption of normally distributed errors. Results indicate that indirect derivation can lead to error characterizations of the transformed variables as accurate or more accurate than direct estimation of the error distributions.
Resource Type
Date Available
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Subject
Rights Statement
Language
Digitization Specifications
  • PDF derivative scanned at 300 ppi (256 B&W), using Capture Perfect 3.0, on a Canon DR-9080C. CVista PdfCompressor 3.1 was used for pdf compression and textual OCR.
Replaces
Additional Information
  • description.provenance : Submitted by Anna Opoien (aoscanner@gmail.com) on 2009-01-09T00:46:57Z No. of bitstreams: 1 Canavan_Sean_2001.pdf: 1252693 bytes, checksum: f6085f62d9b3970e10afa0396f7bc838 (MD5)
  • description.provenance : Approved for entry into archive by Linda Kathman(linda.kathman@oregonstate.edu) on 2009-01-14T18:38:43Z (GMT) No. of bitstreams: 1 Canavan_Sean_2001.pdf: 1252693 bytes, checksum: f6085f62d9b3970e10afa0396f7bc838 (MD5)
  • description.provenance : Made available in DSpace on 2009-01-14T18:39:38Z (GMT). No. of bitstreams: 1 Canavan_Sean_2001.pdf: 1252693 bytes, checksum: f6085f62d9b3970e10afa0396f7bc838 (MD5)
  • description.provenance : Approved for entry into archive by Linda Kathman(linda.kathman@oregonstate.edu) on 2009-01-14T18:39:37Z (GMT) No. of bitstreams: 1 Canavan_Sean_2001.pdf: 1252693 bytes, checksum: f6085f62d9b3970e10afa0396f7bc838 (MD5)

Relationships

In Administrative Set:
Last modified: 08/02/2017

Downloadable Content

Download PDF
Citations:

EndNote | Zotero | Mendeley

Items