1 / 15

Challenges in Applying Nearest Neighbour Methods

Challenges in Applying Nearest Neighbour Methods. Valerie LeMay and Ian Moss. Western Mensurationists Meeting, June 21-23, 2009, Vancouver, Washington. Challenges and Opportunities in Applying Nearest Neighbour Methods. Valerie LeMay and Ian Moss.

cascata-firbis
Download Presentation

Challenges in Applying Nearest Neighbour Methods

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Challenges in Applying Nearest Neighbour Methods Valerie LeMay and Ian Moss Western Mensurationists Meeting, June 21-23, 2009, Vancouver, Washington

  2. Challenges and Opportunities in Applying Nearest Neighbour Methods Valerie LeMay and Ian Moss Western Mensurationists Meeting, June 21-23, 2009, Vancouver, Washington

  3. Dr. Albert Stage

  4. Outline • The Good • The Bad • The

  5. Outline • The Good • The Bad • The Ugly • Million Dollar Baby (Quiz at the End)

  6. The Good: Multivariate Reference Data, X’s and Y’s → Calculate Variable-Space Distance using X’s Target Observation, X’s only Use Y values (or averages) from selected reference observation(s) as Estimates for the target observation Select one or more neighbours that have similar X values (Small distance metric)

  7. The Good: Integrate Data Sources Across Scales For forest Inventory, X’s can be LeMay and Moss

  8. The Bad: Spatial Registration Difficulties

  9. The Bad: Differences in Spatial Extents

  10. The Bad: Too Many Choices? • Choice of weights for auxilliary variables to find nearest neighbours as this affects outcome • How many neighbours to average in getting estimates of the Y’s for a target? • Which X variables to use, which relates to sources of data for the X’s? • Pre-stratification by X variables? • When are results good enough to use?

  11. The Bad: Estimators • Bias Corrections and Variance Estimators? Single Y variable Multiple Y’s given correlations among Y’s?

  12. The Ugly • Complexity of the y-variables (e.g. tree list of dbh and species; biomass of components) • Poor relationships between available X-variables and Y-variables of interest (e.g., stems per ha as the Y variable) • Scarcity of reference data in terms of number of observations and ranges of X-variables, relative to the number of target observations • Processing times: many Y’s and many target observations = very long processing • “Smoothing” of results for mapping when pixel-type data are used

  13. Million Dollar Baby • Research on estimators that have good statistical properties are being developed • Possible gains with new and more available remotely sensed imagery • Better GPS to spatially register data sources • New algorithms for smoothing using spatial clustering • Increased computing resources • Finally – we can link single-tree growth and yield models to the inventory

  14. The Quiz • Q1. What did the list of topics of the outline have in common? (Oral) • Q2. What was uniquely different about one of the items in the list of topics? Name two differences. • Q3. Graduate students only: What is the main take home message? (May be not obvious and remember to “think like a graduate student”) • Q4. What is the nearest neighbour for the target given in the next slide? Why? (You may think outside the box!) LeMay and Moss

  15. Target 2 4 3 5 7 1 6

More Related