web service for toxicant trigger valuation · 2014-02-18 · web service for toxicant trigger...
TRANSCRIPT
Web Service for toxicant
trigger valuation
SUPERVISOR: DR. WARREN JIN
PRESENTER : SHICHAO DONG(U5135470)
Individual Project Courses in the Research School of Computer Science
Overview
Experiment
On
Models
Web
ServiceModification
Overview
Web
Service
BurrliOZ
NEC
Motivation
Extension
Outline BurrliOZ
Model
Experiment
Web Service
NEC
Model
Experiment
Web Service
Discussion
Future Work
Conclusion
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Model
Burr Type III Log-Logistic
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Model
Burr Type III
𝐹 𝑥; 𝜃 = 1 −1
(1+𝑥𝑐)𝑘, x > 0, c > 0, k > 0
𝐹 𝑥; 𝜃 = [1 + (𝑏
𝑥−𝑎)𝑐]−𝑘 , c > 0, k > 0
Original :
Re-rewrite :
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Model
𝐹𝐿𝐿 𝑥 =1
1 + (𝑒𝜇
𝑥)1/𝜃
Distribution Function :
Log-Logistic
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Test
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Test
Features:
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
WebService
Shiny fromBurrliOz2
R Language
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
WebService
Shiny
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
WebService
fromBurrliOz2
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
WebService
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Data Summary
A statistical software package to help environmental managers to figure out trigger values of toxicant
WebService
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Setting
WebService
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Report
Download
Report
A statistical software package to help environmental managers to figure out trigger values of toxicant
WebService
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
Plot
WebService
A statistical software package to help environmental managers to figure out trigger values of toxicant
BurrliOZ
BurrliOZ NEC Discussion Future Work Conclusion
A statistical software package to help environmental managers to figure out trigger values of toxicant
WebService
Setting
NEC
BurrliOZ NEC Discussion Future Work Conclusion
NEC Model
E 𝑌𝑖; 𝑥𝑖 = μ = 𝛼 exp[−𝛽 𝑥𝑖 − 𝛾 𝐼(𝑥𝑖 − 𝛾)]
A Bayesian model for the NEC :
With
𝐼 𝑧 = 1, 𝑧 ≥ 00, 𝑧 ≤ 0
No effect concentration
Model
NEC
BurrliOZ NEC Discussion Future Work Conclusion
Model
E 𝑌𝑖; 𝑥𝑖 = μ = 𝛼 exp[−𝛽 𝑥𝑖 − 𝛾 𝐼(𝑥𝑖 − 𝛾)]
A Bayesian model for the NEC :
With
𝐼 𝑧 = 1, 𝑧 ≥ 00, 𝑧 ≤ 0
α is the response at zero/low − dose concentrationsβ controls the rate of decay in the responseγ is the NEC(no effect concentration)
No effect concentration
NEC
BurrliOZ NEC Discussion Future Work Conclusion
No effect concentration
WinBUGS OpenBUGS
BUGS = Bayesian inference Using Gibbs Sampling
Test
NEC
BurrliOZ NEC Discussion Future Work Conclusion
No effect concentration
Test
NEC
BurrliOZ NEC Discussion Future Work Conclusion
Test
WinBUGS OpenBUGS
No effect concentration
WinBUGS OpenBUGS
Result Repeatable No Yes
Code Running No Yes
Running on Linux No Yes
Still update No Yes
Same function Yes yes
NEC
BurrliOZ NEC Discussion Future Work Conclusion
R codeShinyUser
Interface Server Core
No effect concentration
WebService
OpenBUGSLibrary
NEC
BurrliOZ NEC Discussion Future Work Conclusion
No effect concentration
WebService
NEC
BurrliOZ NEC Discussion Future Work Conclusion
No effect concentration
WebService
NEC
BurrliOZ NEC Discussion Future Work Conclusion
WebService
No effect concentration
Discussion
Cannot repeat the result
Comparison
Extension
BurrliOZ NEC Discussion Future Work Conclusion
Discussion
BurrliOZ NEC Discussion Future Work Conclusion
Possible reason: has not updated for yearsOur solution: Using OpenBUGS
Discussion
BurrliOZ NEC Discussion Future Work Conclusion
Parameters
Mean Standard deviation
P2.5 P50 P97.5
Alpha 0.95 0.01 0.92 0.95 0.97
Beta 1188 3139 17 107 12350
Gamma 0.95 0.08 0.01 0.86 1.11
Possible Reasons: Typo Error, Software problemSolution : continue using OpenBUGS
Discussion
BurrliOZ BurrliOZ(WEB) NEC NEC(WEB)
Interface Yes Yes No Yes
OS Windows Any Any Any
Platform Computer Computer,mobile device
Computer Computer,mobile device
Software .net library Web Browser WinBUGS Web Browser
Programming No No Yes NO
BurrliOZ NEC Discussion Future Work Conclusion
Comparison
More flexible and accessible usage
Discussion
Features BurrliOZ BurrliOZ(WEB) NEC NEC(WEB)
Loading data
Models 2 3 1 1
Priors distribution NA NA 1 2
Output result
BurrliOZ NEC Discussion Future Work Conclusion
Extension Features
More features and more comparison
Future Work
update the BurrliOZ package;
improve the performance of BurrliOZ;
more possible priors distributions in NEC.
rewriting the front-end by using JavaScript;
testing new possible models;
comparing NOEC. vs NEC or other models
using different tools and packages;
using the same way in different fields
BurrliOZ NEC Discussion Future Work Conclusion
Conclusion
Revisiting and experimenting Shao’s and Fox’s work, and we find there seems to be different from Fox’s result with ours, we try to explain the difference and give our own solutions;
implementing BurrliOZ into web service;
implementing NEC into web service;
providing figures with published quality
extending Fox’s work via providing more flexibility in the web service like normal distribution and Weibull distribution
BurrliOZ NEC Discussion Future Work Conclusion
ExperimentWeb
ServiceExtension
Reference [1] Tom Aldenberg and Wouter Slob. Confidence limits for hazardous concentrations based on logistically distributed noec toxicity data. Ecotoxicology and
Environmental Safety, 25(1):48–63, 1993.
[2] Christophe Andrieu, Nando De Freitas, Arnaud Doucet, and Michael I Jordan. An introduction to mcmc for machine learning. Machine learning, 50(1-2):5–43, 2003.
[3] Kenneth E Biesinger, Leroy E Anderson, and John G Eaton. Chronic effects of inorganic and organic mercury ondaphnia magna: Toxicity, accumulation, and loss. Archives of environmental contamination and toxicology, 11(6):769–774,1982.
[4] MRC Biostatistics. Background to bugs. http://www.mrc-bsu.cam.ac.uk/bugs/.
[5] Irving W Burr. Cumulative frequency functions. The Annals of Mathematical Statistics, 13(2):215–232, 1942.
[6] CSIRO. Burrlioz 2.0 updates. http://www.cmis.csiro.au/Envir/Burrlioz/.
[7] CSIRO. Burrlioz statistical software: a flexible approach to species protection.http://www.csiro.au/Outcomes/Environment/Australian-Landscapes/BurrliOZ.aspx.
[8] A.J. Culyer. The Dictionary of Health Economics. Elgar Original Reference Series. Edward Elgar Publishing, Incorporated, 2005.
[9] David R Fox. A bayesian approach for determining the no effect concentration and hazardous concentration in ecotoxicology. Ecotoxicology and Environmental Safety, 73(2):123–131, 2010.
[10] Jeffrey H Gove, Mark J Ducey, William B Leak, and Lianjun Zhang. Rotated sigmoid structures in managed uneven-aged northern hardwood stands: a look at the burr type iii distribution. Forestry, 81(2):161–176, 2008.
[11] SR Lindsay, GR Wood, and RC Woollons. Modelling the diameter distribution of forest stands using the burr distribution. Journal of Applied Statistics, 23(6):609–620, 1996.
[12] Ana M Pires, João A Branco, Ana Picado, and Elsa Mendonça. Models for the estimation of a no effect concentration. Environmetrics, 13(1):15–27, 2002.
[13] r project. Introduction to r. http://www.r-project.org/.
[14] B.K. Shah and P.H. Dave. A note on log-logistic distribution. Journal of the MS University of Baroda (Science Number), 12:15–20, 1963.
[15] Quanxi Shao. Estimation for hazardous concentrations based on noec toxicity data: an alternative approach. Environmetrics, 11(5):583–595, 2000.
[16] Pandu R Tadikamalla. A look at the burr and related distributions. International Statistical Review/Revue Internationale de Statistique, pages 337–344, 1980.
[17] Pandu R Tadikamalla and Norman L Johnson. Systems of frequency curves generated by transformations of logistic variables. Biometrika, 69(2):461–465, 1982.
[18] Neal Thomas. Changes between winbugs and openbugs. http://www. openbugs.info/w.cgi/OpenVsWin.
[19] Neal Thomas. How it works. http://www.openbugs.info/w/.
[20] Waloddi Weibull et al. A statistical distribution function of wide applicability. Journal of applied mechanics, 18(3):293–297, 1951.
Thank you !
Q & A