ECE 695A Lecture 33: Model Selection/Goodness of Fit

By Muhammad Alam

Electrical and Computer Engineering, Purdue University, West Lafayette, IN

View Presentation

Additional materials available (4)

Licensed under Creative Commons BY-NC-SA 2.5.

Category

Online Presentations

Published on

Abstract

Outline:

  • The problem of matching data with theoretical distribution
  • Parameter extractions: Moments, linear regression, maximum likelihood
  • Goodness of fit: Residual, Pearson, Cox, Akika
  • Conclusion

Cite this work

Researchers should cite this work as follows:

  • Muhammad Alam (2013), "ECE 695A Lecture 33: Model Selection/Goodness of Fit," https://nanohub.org/resources/17615.

    BibTex | EndNote

Time

Location

EE 226, Purdue University, West Lafayette, IN

Tags

  1. course lecture
  2. experiments
  3. nanoelectronics
  4. reliability
ECE 695A Lecture 33: Model Selection/Goodness of Fit
  • Lecture 33: Model Selection/Goodness of Fit 1. Lecture 33: Model Selection/Go… 0
    00:00/00:00
  • copyright 2013 2. copyright 2013 161.99532866199533
    00:00/00:00
  • Outline 3. Outline 163.29662996329662
    00:00/00:00
  • Problem of matching the moments 4. Problem of matching the moment… 276.343009676343
    00:00/00:00
  • (1) Linear regression: balanced errors 5. (1) Linear regression: balance… 359.09242575909246
    00:00/00:00
  • Uncertainty in regression coefficients 6. Uncertainty in regression coef… 558.692025358692
    00:00/00:00
  • Methods of least squares for weibull 7. Methods of least squares for w… 769.70303636970311
    00:00/00:00
  • Fitting of physical models: challenges 8. Fitting of physical models: ch… 847.38071404738071
    00:00/00:00
  • Method of correlation coefficient 9. Method of correlation coeffici… 937.80447113780451
    00:00/00:00
  • (2) Fisher's Maximum Likelihood Method 10. (2) Fisher's Maximum Likelihoo… 1188.2882882882884
    00:00/00:00
  • Example: origin of least square method 11. Example: origin of least squar… 1500.1668335001668
    00:00/00:00
  • Example (continued) 12. Example (continued) 1590.7907907907909
    00:00/00:00
  • Example: MLE estimator for one-parameter distribution 13. Example: MLE estimator for one… 1648.7821154487822
    00:00/00:00
  • Example: MLE estimator for Weibull 14. Example: MLE estimator for Wei… 1771.0377043710378
    00:00/00:00
  • HW: MLE for Log-Normal 15. HW: MLE for Log-Normal 1898.8321654988322
    00:00/00:00
  • Outline 16. Outline 1948.9823156489824
    00:00/00:00
  • (1) Goodness of Fit: First check visually 17. (1) Goodness of Fit: First che… 1992.5925925925926
    00:00/00:00
  • (2) Goodness of Fit: Residual method 18. (2) Goodness of Fit: Residual … 2031.0977644310979
    00:00/00:00
  • (3) Goodness of fit: Q-Q Method 19. (3) Goodness of fit: Q-Q Metho… 2137.4040707374043
    00:00/00:00
  • Q-Q Method: An example 20. Q-Q Method: An example 2330.8641975308642
    00:00/00:00
  • Q-Q method: an example 21. Q-Q method: an example 2342.4424424424424
    00:00/00:00
  • (4) Goodness of Fit: Cox-Oakes measure 22. (4) Goodness of Fit: Cox-Oakes… 2385.251918585252
    00:00/00:00
  • (5) Kolmogorov-smirnov algorithm 23. (5) Kolmogorov-smirnov algorit… 2728.4951618284954
    00:00/00:00
  • Example: Kolmogorov-Smirnov Test 24. Example: Kolmogorov-Smirnov Te… 2879.6463129796466
    00:00/00:00
  • (6) Pearson χ2 – test algorithm 25. (6) Pearson χ2 – test algor… 3195.4954954954956
    00:00/00:00
  • A famous example: Schon story 26. A famous example: Schon story 3341.775108441775
    00:00/00:00
  • Example: Shunt statistics is log-normal 27. Example: Shunt statistics is l… 3536.2028695362028
    00:00/00:00
  • Parameter number vs. goodness of fit 28. Parameter number vs. goodness … 3595.2952952952955
    00:00/00:00
  • Conclusions 29. Conclusions 3765.3653653653655
    00:00/00:00
  • References 30. References 3809.60960960961
    00:00/00:00
Pop Out