1 / 18

Uncertainty Estimation

Uncertainty Estimation. Simon Cousens and Richard Silverwood. Overview. an implementation/coding issue with the Loess approach accounting for within-source correlation increasing uncertainty with increasing sources(?). Implementation issue.

ismet
Download Presentation

Uncertainty Estimation

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. Uncertainty Estimation Simon Cousens and Richard Silverwood

  2. Overview • an implementation/coding issue with the Loess approach • accounting for within-source correlation • increasing uncertainty with increasing sources(?)

  3. Implementation issue • Murray at al. describe incorporation of two sources of uncertainty – random noise in data points through resampling parameter values using point estimates and variance-covariance matrix - (one aspect of) model uncertainty through use of a range of α values • Wardlaw subsequently noted surprisingly narrow uncertainty ranges for Belarus, Congo, Cote d’Ivoire, Ukraine

  4. Example: Congo

  5. Example: Belarus

  6. Accounting for within source correlation • Loess and spline approaches have assumed independence of data points • In practice multiple data points used in analysis derived from a single source (e.g. DHS) and assumption of independence likely to be violated One approach is to fit a multi-level model. Loess model: log(yij) = β0j + β1xij +β2zij + eij where β0j = β0 + uj and uj ~ N(0,σu)

  7. Results In general, little change in point estimates (e.g. 2010)

  8. Results Uncertainty ranges within period of observed data increase (e.g. 1990)

  9. Results Pattern less clear outside period of observed data (e.g. 2010)

  10. Some specific country examples Costa Rica

  11. Some specific country examples Sierra Leone

  12. Increasing uncertainty with increasing sources? Eritrea

  13. Eritrea

  14. Eritrea

  15. Comoros

  16. The Gambia

  17. Bangladesh

  18. Bangladesh No obvious “asymtotic” pattern – is this useable in any way?

More Related