1 Some Metrological Aspects of Ordinal Quality Data Treatment *Emil Bashkansky Tamar Gadrich ORT Braude College of Engineering, Israel  ENBIS-11 Coimbra,

Slides:



Advertisements
Similar presentations
Modeling of Data. Basic Bayes theorem Bayes theorem relates the conditional probabilities of two events A, and B: A might be a hypothesis and B might.
Advertisements

Combining Classification and Model Trees for Handling Ordinal Problems D. Anyfantis, M. Karagiannopoulos S. B. Kotsiantis, P. E. Pintelas Educational Software.
Chapter 6 Sampling and Sampling Distributions
2 – In previous chapters: – We could design an optimal classifier if we knew the prior probabilities P(wi) and the class- conditional probabilities P(x|wi)
Integration of sensory modalities
The adjustment of the observations
Slide 1 EE3J2 Data Mining EE3J2 Data Mining Lecture 10 Statistical Modelling Martin Russell.
G. Cowan Lectures on Statistical Data Analysis 1 Statistical Data Analysis: Lecture 10 1Probability, Bayes’ theorem, random variables, pdfs 2Functions.
0 Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John.
Statistical Inference Chapter 12/13. COMP 5340/6340 Statistical Inference2 Statistical Inference Given a sample of observations from a population, the.
Evaluating Hypotheses
Pattern Classification All materials in these slides were taken from Pattern Classification (2nd ed) by R. O. Duda, P. E. Hart and D. G. Stork, John Wiley.
Part III: Inference Topic 6 Sampling and Sampling Distributions
Copyright © Cengage Learning. All rights reserved. 6 Point Estimation.
Market Risk VaR: Historical Simulation Approach
MEASURING PROCESS QUALITY ON AN ORDINAL SCALE BASIS E. Bashkansky, T.Gadrich Industrial Engineering & Management Department E.Godik Software Engineering.
Effectiveness of a Product Quality Classifier Dr. E. Bashkansky, Dr. S. Dror, Dr. R. Ravid Industrial Eng. & Management, ORT Braude College, Karmiel, Israel.
1 Presentation Outline Introduction Objective Sorting Probability and Loss Matrices The Proposed Model Analysis Of Some Loss Functions Case Study Redundancy.
Lecture II-2: Probability Review
METU Informatics Institute Min 720 Pattern Classification with Bio-Medical Applications PART 2: Statistical Pattern Classification: Optimal Classification.
EE513 Audio Signals and Systems Statistical Pattern Classification Kevin D. Donohue Electrical and Computer Engineering University of Kentucky.
Digital Camera and Computer Vision Laboratory Department of Computer Science and Information Engineering National Taiwan University, Taipei, Taiwan, R.O.C.
Copyright © Cengage Learning. All rights reserved. 8 Tests of Hypotheses Based on a Single Sample.
Inference for the mean vector. Univariate Inference Let x 1, x 2, …, x n denote a sample of n from the normal distribution with mean  and variance 
Magister of Electrical Engineering Udayana University September 2011
Classification (Supervised Clustering) Naomi Altman Nov '06.
Estimation Basic Concepts & Estimation of Proportions
Principles of Pattern Recognition
Digital Camera and Computer Vision Laboratory Department of Computer Science and Information Engineering National Taiwan University, Taipei, Taiwan, R.O.C.
ECE 8443 – Pattern Recognition LECTURE 03: GAUSSIAN CLASSIFIERS Objectives: Normal Distributions Whitening Transformations Linear Discriminants Resources.
Statistics for Managers Using Microsoft Excel, 5e © 2008 Pearson Prentice-Hall, Inc.Chap 8-1 Statistics for Managers Using Microsoft® Excel 5th Edition.
ECE 8443 – Pattern Recognition LECTURE 07: MAXIMUM LIKELIHOOD AND BAYESIAN ESTIMATION Objectives: Class-Conditional Density The Multivariate Case General.
Classification Techniques: Bayesian Classification
Lecture 4: Statistics Review II Date: 9/5/02  Hypothesis tests: power  Estimation: likelihood, moment estimation, least square  Statistical properties.
CSSE463: Image Recognition Day 11 Lab 4 (shape) tomorrow: feel free to start in advance Lab 4 (shape) tomorrow: feel free to start in advance Test Monday.
: Chapter 3: Maximum-Likelihood and Baysian Parameter Estimation 1 Montri Karnjanadecha ac.th/~montri.
Statistical Decision Theory Bayes’ theorem: For discrete events For probability density functions.
B. Neidhart, W. Wegscheider (Eds.): Quality in Chemical Measurements © Springer-Verlag Berlin Heidelberg 2000 U. PyellBasic Course Experiments to Demonstrate.
Pairwise Preference Regression for Cold-start Recommendation Speaker: Yuanshuai Sun
Variables It is very important in research to see variables, define them, and control or measure them.
Chapter 20 Classification and Estimation Classification – Feature selection Good feature have four characteristics: –Discrimination. Features.
Point Estimation of Parameters and Sampling Distributions Outlines:  Sampling Distributions and the central limit theorem  Point estimation  Methods.
CSSE463: Image Recognition Day 11 Due: Due: Written assignment 1 tomorrow, 4:00 pm Written assignment 1 tomorrow, 4:00 pm Start thinking about term project.
Logistic Regression Saed Sayad 1www.ismartsoft.com.
1 Chapter 8: Model Inference and Averaging Presented by Hui Fang.
Part 4: Contextual Classification in Remote Sensing * There are different ways to incorporate contextual information in the classification process. All.
Classifiers!!! BCH364C/391L Systems Biology / Bioinformatics – Spring 2015 Edward Marcotte, Univ of Texas at Austin.
Establishing by the laboratory of the functional requirements for uncertainty of measurements of each examination procedure Ioannis Sitaras.
MathematicalMarketing Slide 3c.1 Mathematical Tools Chapter 3: Part c – Parameter Estimation We will be discussing  Nonlinear Parameter Estimation  Maximum.
Logic of Hypothesis Testing
Why Stochastic Hydrology ?
Classifiers!!! BCH339N Systems Biology / Bioinformatics – Spring 2016
CSSE463: Image Recognition Day 11
Classification with Perceptrons Reading:
Classification Techniques: Bayesian Classification
REMOTE SENSING Multispectral Image Classification
9 Tests of Hypotheses for a Single Sample CHAPTER OUTLINE
Basic Statistical Terms
Evaluating Classifiers (& other algorithms)
Where did we stop? The Bayes decision rule guarantees an optimal classification… … But it requires the knowledge of P(ci|x) (or p(x|ci) and P(ci)) We.
Integration of sensory modalities
EE513 Audio Signals and Systems
5.2 Least-Squares Fit to a Straight Line
Bayesian Classification
Parametric Methods Berlin Chen, 2005 References:
Learning From Observed Data
CSSE463: Image Recognition Day 11
Mathematical Foundations of BME Reza Shadmehr
CSSE463: Image Recognition Day 11
Evaluating Classifiers
Presentation transcript:

1 Some Metrological Aspects of Ordinal Quality Data Treatment *Emil Bashkansky Tamar Gadrich ORT Braude College of Engineering, Israel  ENBIS-11 Coimbra, Portugal,  September 2011, 11:50 – 13:20

2 Presentation Outline-stage I Why revision of MC for ordinal measurements is needed? Error Binary caseGeneral case Uncertainty General caseBinary case

3 Presentation Outline-stage II Main metrological characteristics AccuracyPrecision RepeatabilityReproducibility

4 Presentation Outline-stage III Repeated measurements Binary caseGeneral case

5 Examples of ordinal scale usage DAILY LIFE MEDICINE QUALITY MANAGEMENT Engineering Sports results: a win, tie, loss Voting results: pro, against, abstain Academic ranks … Rankin score (RS) - level of disability following a stroke Side effect severity … Quality level estimation and sorting Customer satisfaction surveys Ratings of wine colour, aroma and taste FMECA … The Mohs scale of mineral hardness Dry-chemistry dipsticks (e.g., urine test) The Beaufort wind force scale....

6 Why revision of MC for ordinal measurements is needed? * ISO/IEC Guide 99: International vocabulary of metrology — “Basic and general concepts and associated terms (VIM)”

7 Classic continual: The probability density function pdf (Y/X) of receiving result Y, given the true value of the measurand X, in it's simplest form: pdf (Y/X) = Normal (X+bias, ) Ordinal: The conditional probabilities that an object will be classified as level j, given that its actual/true level is i. Error description

8 Error -free ordinal measurement

9 Error –binary case

10 Some examples

11 Uncertainty-general case The likelihood that a measured level j is received, whereas the true level is i

12 Uncertainty matrix- binary case

13 Inaccuracy- Error matrix :

14 Effectiveness vs. Accuracy Effectiveness of the measurement system Bashkansky E, Dror S, Ravid R, Grabov P (2007) Effectiveness of a Product Quality Classifier. Quality Engineering 19(3):

15 Precision (the closeness of agreement between independent test results obtained under stipulated conditions) Repeatability (same conditions) Reproducibility (different conditions)

16 Repeatability Blair & Lacy (2000)

17 Repeatability - the expected cumulative frequency of data/items classified up to the k-th category, given that its actual/true level is i

18 - the expected cumulative frequency of items belonging up to the k-th category after measurement ORDANOVA: DECOMPOSITION OF TOTAL DISPERSION AFTER MEASUREMENT/CLASSIFICATION

19 Reference standard (known/unknown) Measurement system A Measurement system B

20 Some definitions - conditional joint probability of sorting the measured object to the a-th level by the first MS (called A), and the b-th level by the second MS (called B), given the actual/true category i

21 A & B MSs classification matrices

22 Joint probability matrix p i - the probability that an object being measured relates to category i, ( ) - the joint probability of sorting the item as a by the first measurement system (A) and b by the second measurement system (B).

23 MODIFIED KAPPA MEASURE OF AGREEMENT m When a half of all items are correctly classified:

24 1. Reproducibility - reference standard is known

25 2. Reproducibility - reference standard is unknown

26 Binary case example

27 QUALITY CATEGORY SOLUBLE SOLIDS CONTENT (SSC) TITRATABLE ACIDITY (TA) PH TOTAL SUGARS MASS FRACTIONS SKIN COLOR ("A" VALUE) FLESH FIRMNESS MASS (%) (G/KG) (N)(G) HIGH (TYPE 1) >105 MEDIUM (TYPE 2) LOW (TYPE 3) <12>1.1<3.6<50>25>55<85 Typical relation between quality level and commonly used chemical/physical features for yellow-flesh nectarines

28 weighted total kappa equals Ternary scale example (fruit quality classification)

29 Repeated measurements-binary case (n = n 1 +n 2 )

30 Repeated measurements-binary case αβ CUT POINT

31 Binary case example

32 LIKELIHOODS (given the true value one or two) vs. n 1 /n LIKELIHOODS (given the true value one or two) vs. n 1 /n

33 Repeated measurements-general case Let's consider arbitrary ordinal scale with m categories and suppose, that n repeated measurements of the same object were performed resulting in vector: (n=n 1 +n 2 +…+n m )

34 Repeated measurements-general case The maximum likelihood estimation must be made in favor of such, most plausible i, that maximizes the scalar product:

35 General case example-single measurement

36 General case example: after 10 repititions

37 Binary case free access calculator

38SUMMARY 1.On an ordinal measurement scale the essential for evaluating the error, repeatability and uncertainty of the measurement result base knowledge must be the classification/measurement matrix. Given this matrix, authors introduced a way to calculate the classification/measurement system’s accuracy, precision (repeatability & reproducibility) and uncertainty matrix. 2.In order to estimate comparability and equivalence between measurement results received on an ordinal scale basis, the modified kappa measure is suggested. Three of the most suitable usages of the measure were thoroughly analyzed. The advantage of the proposed measure vs. the traditional one lies in the fact that the former follows the superposition principle: the total measure equals the weighted sum of partial measures for every ordinal category. 3.As it is well known, repeated measurements may improve the quality of the measurement result. When decisions are ML based, one can find how many repetitions are necessary in order to achieve the desired accuracy level using the algorithm suggested by the authors,.

39 1.Bashkansky E., Dror S., Ravid R., Grabov P. (2007), “Effectiveness of a Product Quality Classifier”, Quality Engineering, vol. 19, issue 3, pp Bashkansky E., Gadrich T., (2010) “Some Metrological Aspects of Ordinal Measurements”, Accreditation and Quality Assurance, vol. 15, pp Bashkansky E., Gadrich T., Knani D., (2011) “Some metrological aspects of the comparison between two ordinal measuring systems”, Accreditation and Quality Assurance, vol. 16, pp

40

41