A Study of the Reproducibility of Measurements with HUR Leg Extension/Curl Research Line

Similar documents
CHAPTER VI Statistical Analysis of Experimental Data

best estimate (mean) for X uncertainty or error in the measurement (systematic, random or statistical) best

Summary of the lecture in Biostatistics

Lecture Notes Types of economic variables

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Simulation Output Analysis

Econometric Methods. Review of Estimation

Continuous Distributions

Functions of Random Variables

Descriptive Statistics

Lecture 7. Confidence Intervals and Hypothesis Tests in the Simple CLR Model

Class 13,14 June 17, 19, 2015

Lecture 3 Probability review (cont d)

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Lecture 1 Review of Fundamental Statistical Concepts

Chapter 5 Properties of a Random Sample

Quantitative analysis requires : sound knowledge of chemistry : possibility of interferences WHY do we need to use STATISTICS in Anal. Chem.?

Chapter 8. Inferences about More Than Two Population Central Values

STA 108 Applied Linear Models: Regression Analysis Spring Solution for Homework #1

Chapter 8: Statistical Analysis of Simulated Data

Evaluation of uncertainty in measurements

Special Instructions / Useful Data

Estimation of Stress- Strength Reliability model using finite mixture of exponential distributions

hp calculators HP 30S Statistics Averages and Standard Deviations Average and Standard Deviation Practice Finding Averages and Standard Deviations

f f... f 1 n n (ii) Median : It is the value of the middle-most observation(s).

ENGI 4421 Joint Probability Distributions Page Joint Probability Distributions [Navidi sections 2.5 and 2.6; Devore sections

Chapter 13, Part A Analysis of Variance and Experimental Design. Introduction to Analysis of Variance. Introduction to Analysis of Variance

Mean is only appropriate for interval or ratio scales, not ordinal or nominal.

X ε ) = 0, or equivalently, lim

Module 7. Lecture 7: Statistical parameter estimation

MEASURES OF DISPERSION

Third handout: On the Gini Index

Introduction to local (nonparametric) density estimation. methods

Lecture 2 - What are component and system reliability and how it can be improved?

Ordinary Least Squares Regression. Simple Regression. Algebra and Assumptions.

Lecture 3. Sampling, sampling distributions, and parameter estimation

( ) = ( ) ( ) Chapter 13 Asymptotic Theory and Stochastic Regressors. Stochastic regressors model

STATISTICAL PROPERTIES OF LEAST SQUARES ESTIMATORS. x, where. = y - ˆ " 1

2.28 The Wall Street Journal is probably referring to the average number of cubes used per glass measured for some population that they have chosen.

Random Variables and Probability Distributions

Midterm Exam 1, section 1 (Solution) Thursday, February hour, 15 minutes

ESS Line Fitting

Law of Large Numbers

Point Estimation: definition of estimators

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA

ENGI 3423 Simple Linear Regression Page 12-01

Section l h l Stem=Tens. 8l Leaf=Ones. 8h l 03. 9h 58

Bayesian Inferences for Two Parameter Weibull Distribution Kipkoech W. Cheruiyot 1, Abel Ouko 2, Emily Kirimi 3

STA 105-M BASIC STATISTICS (This is a multiple choice paper.)

Analysis of Variance with Weibull Data

Module 7: Probability and Statistics

Chapter 14 Logistic Regression Models

Parameter, Statistic and Random Samples

Simple Linear Regression

Statistics. Correlational. Dr. Ayman Eldeib. Simple Linear Regression and Correlation. SBE 304: Linear Regression & Correlation 1/3/2018

PROPERTIES OF GOOD ESTIMATORS

9 U-STATISTICS. Eh =(m!) 1 Eh(X (1),..., X (m ) ) i.i.d

Simple Linear Regression

Multiple Regression. More than 2 variables! Grade on Final. Multiple Regression 11/21/2012. Exam 2 Grades. Exam 2 Re-grades

is the score of the 1 st student, x

BAYESIAN INFERENCES FOR TWO PARAMETER WEIBULL DISTRIBUTION

Multivariate Transformation of Variables and Maximum Likelihood Estimation

Lecture 8: Linear Regression

A Combination of Adaptive and Line Intercept Sampling Applicable in Agricultural and Environmental Studies

GOALS The Samples Why Sample the Population? What is a Probability Sample? Four Most Commonly Used Probability Sampling Methods

Bayes Estimator for Exponential Distribution with Extension of Jeffery Prior Information

The equation is sometimes presented in form Y = a + b x. This is reasonable, but it s not the notation we use.

Statistics MINITAB - Lab 5

ENGI 4421 Propagation of Error Page 8-01

Bootstrap Method for Testing of Equality of Several Coefficients of Variation

Correlation and Simple Linear Regression

THE ROYAL STATISTICAL SOCIETY GRADUATE DIPLOMA

Lecture Notes to Rice Chapter 5

Chapter 2 - Free Vibration of Multi-Degree-of-Freedom Systems - II

A New Family of Transformations for Lifetime Data

BIOREPS Problem Set #11 The Evolution of DNA Strands

1 Mixed Quantum State. 2 Density Matrix. CS Density Matrices, von Neumann Entropy 3/7/07 Spring 2007 Lecture 13. ψ = α x x. ρ = p i ψ i ψ i.

X X X E[ ] E X E X. is the ()m n where the ( i,)th. j element is the mean of the ( i,)th., then

On Fuzzy Arithmetic, Possibility Theory and Theory of Evidence

C-1: Aerodynamics of Airfoils 1 C-2: Aerodynamics of Airfoils 2 C-3: Panel Methods C-4: Thin Airfoil Theory

Objectives of Multiple Regression

Chapter 13 Student Lecture Notes 13-1

Multiple Linear Regression Analysis

Lecture 4 Sep 9, 2015

ρ < 1 be five real numbers. The

{ }{ ( )} (, ) = ( ) ( ) ( ) Chapter 14 Exercises in Sampling Theory. Exercise 1 (Simple random sampling): Solution:

i 2 σ ) i = 1,2,...,n , and = 3.01 = 4.01

= 1. UCLA STAT 13 Introduction to Statistical Methods for the Life and Health Sciences. Parameters and Statistics. Measures of Centrality

SPECIAL CONSIDERATIONS FOR VOLUMETRIC Z-TEST FOR PROPORTIONS

Lecture 12 APPROXIMATION OF FIRST ORDER DERIVATIVES

Comparison of Parameters of Lognormal Distribution Based On the Classical and Posterior Estimates

BAL-001-AB-0a Real Power Balancing Control Performance

Linear Regression. Hsiao-Lung Chan Dept Electrical Engineering Chang Gung University, Taiwan

STA302/1001-Fall 2008 Midterm Test October 21, 2008

Lecture 02: Bounding tail distributions of a random variable

PGE 310: Formulation and Solution in Geosystems Engineering. Dr. Balhoff. Interpolation

MULTIDIMENSIONAL HETEROGENEOUS VARIABLE PREDICTION BASED ON EXPERTS STATEMENTS. Gennadiy Lbov, Maxim Gerasimov

Qualifying Exam Statistical Theory Problem Solutions August 2005

Chapter 5 Elementary Statistics, Empirical Probability Distributions, and More on Simulation

Bounds on the expected entropy and KL-divergence of sampled multinomial distributions. Brandon C. Roy

Transcription:

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) A Study of the Reproducblty of Measuremets wth HUR Leg Eteso/Curl Research Le A mportat property of measuremets s that the results should be the same uder equvalet codtos save for the uavodable statstcal fluctuatos. I physologcal measuremets we do however ever have detcal codtos. Yet, f a perso performs, let us say, a umber of MVC sometrc cotractos wth a short spa of tme (a day or so) uder smlar crcumstaces, such that we ca eglect trag effects, we should epect to obta readgs of the peak force whch do ot vary too much. We wll assume that the measuremet values fluctuate aroud a mea value ad the try to determe the se of the fluctuatos. Our prelmary result s that the measuremet values of peak torque (sometrc ad dyamc tests), ad mamum agular velocty (dyamc test) wll be wth about 0% of the mea value 95% of the cases. Ths s comparable wth results cted for soketc devces whch clam a varato of the measuremet results (peak torque, total work, average force) of the order of 9 4%. The data used the preset study s derved form a prevous study 3. I oe of the schemes employed that study they tested a group of fve wome practsers of aerobcs. Durg oe day of testg the wome performed fve sometrc eteso tests ad fve fleo tests, ad fve sets of dyamc eteso tests wth varyg resstace ad for both legs (left, rght), usg the stadard HUR test meu wth HUR Leg Eteso/Curl. Suppose we have obtaed a seres of values > 0 (,,..., ) by repeatg a measuremet tmes uder smlar codtos. Oe way to decrbe the varato of the result s to form the parameters 4 () ˆ ˆ where the wth a hat deotes the average Mamum volutary cotracto. Quoted (Mäkkö ad Martkae, 000) see ref. below. 3 N Mäkkö, V Martkae, Isometrset ja Dyaamset Mttaukset (Oulu dakossalatos, 000). Ms. 4 Dvdg wth the average value make the varatos for dfferet subjects comparable.

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) () ˆ As a eample of the data we gve the results of the sometrc fleo (curl) tests (left leg, mamum torque Nm) the followg table 5 : 86 77 99 0 5 8 86 0 3 84 8 0 3 4 85 80 89 3 90 79 89 7 Each colum gves the fve results of each subject (N 5). For each colum we caclulate the averages () ad the the 5 5 5 values for the -parameter (). Ths s repeated for rght leg data, ad eteso rght ad left leg data. It s foud that vares betwee about 0. ad 0.. We ca draw a correspodg hstogram for the total data 6 : Fgure Frequecy of the varato of sometrc data (totque) 5 The sometrc test cossts of two measuremets of MVC peak torque ad the better result s take as the represetatve value of the test. The jot agle fleo (curl) measuremets was set to 40 degrees ad the eteso measuremets to 0 degrees. 6 Data cossted of 90 ( 00 5) data pots because two complete tests were dropped. I the dagrams we have used all the -pots though for every test (cosstg of fve measuremets) oe -pot s determed by the four others sce ther sum accordg to the defto () wll be ero. Usg all the - pots gves us smoother curves ad forces the average to be eactly ero.

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) A clearer pcture of the statstcs of the data emerges f we draw the cumulatve dstrbuto of the varato: Fgure Dstrbuto of the varato of the sometrc data Ths shows e.g. that about 85 % of the measuremets the varato s less tha 0.05:.e. the measured value does ot eceed the true mea value by more tha 5 %. The cotous le the dagram s the curve of the ormal dstrbuto (Gauss) wth mea 0 ad stadard devato 0.045. The stadard devato was estmated from s usg (3) s As we ca see from the fgure the gaussa dstrbuto seems to descrbe the dstrbuto of the varato the sometrc measuremet data qute well. The stadard devato of the measuremet data ca be estmated usg 0.045 ad equ (6) the Apped: 0. 05 Thus, based o ths data we ca say that a measuremet of sometrc peak torque wll about 95 percet of the cases be wth ±0 % (0% /) 7 of the epected mea value of such measuremets. 7 The 95% cofdece terval of a ormal dstrbuto s gve by ( -.96, +.96 ). 3

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) That s, f we make two measuremets o a perso ad obta results that dffer by more tha about 4% (0 : 4) 8, the, wth a 95 percet certaty we may epect the dfferece to be sgfcat ad ot wth the bouds of statstcal varato. I the dyamc eteso tests the peak torque ad mamum agular velocty was measured durg MVC at resstaces steps betwee ad 8 bar. Ths was repeated fve tmes. The torque ad velocty data of the dyamc tests was processed the same way as the torque data of the sometrc tests. For peak torque data we used the measuremets at 4 ad 8 bar, ad for velocty data we used measuremets at ad 4 bar resstace (from both rght ad left leg). (Note: ad 4 bar data treated separately show practcally the same dstrbutos whece t makes sese to lump them together.) The dstrbuto of the varato for the peak torque s preseted the followg fgure: Fgure 3 Dstr. of vara. of peak torque of the dyamc tests The ormal dstrbuto draw the fgure has a stadard devato 0.037. For the varato of mamum agular veloctes we obta the dstrbuto: 8 The dfferece of two depedet radom varables wth ormal dstrbuto N(, ) wll have a ormal dstrbuto N(0, ). Formally the se of the sgfcat dfferece of two measuremets m ad m could be epressed as 0. m + m. 4

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) Fgure 4 Dstr. of vara. of ma. agular velocty of the dy. data I ths case the stadard devato of the supermposed ormal dstrbuto curve s 0.04. I fgure 3 ad 4 we see that the agreemet wth the ormal dstrbuto s ot perfect. Ths could suggest that the measuremets does ot eactly follow a ormal dstrbuto aroud a mea value but stead shows a hgher cocetrato aroud the ero varato pot 0. Oe mght speculate though ths data does ot warrat ay far reachg coclusos ths respect (the bumps are pretty much wth the statstcal fluctuatos) - e.g. that the dyamc tests the motorc program actvated also stables the results. Submamal efforts whch perhaps allow for greater cotrol of the moto could also stable the results. To test ths more data would be eeded. The esmato of of the stadard devato usg (3) s also fraught wth some ucertaty. Ideed, f were depedet ormal varables wth the dstrbuto N(0, ), the the sum χ N would obey the Ch-square dstrbuto. If the degrees of freedom N s greater tha 30 the Ch-square dstrbuto s well appromated by a ormal dstrbuto N( N, N ). I our case we obta from by droppg every ffth varable sce accordg to () the sum of the fve measuremets of every test s ero; thus, oly 4 5 are depedet varables. Wth these cosderatos we get N (4/5) 00 80 ad Prob 80 [.96 80 χ 80 +.96 80] 0. 95 Thus the 95% cofdece terval for the estmate of the stadard devato 5

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) s N N N χ wll be.96, 80 +.96 80 whch s (0.85,.5 );.e. the stadard varato s wth 5% of the estmate, where we for ca use the estmate s or (3) whch wll be of the same se (aroud 0.04). The pot of ths s oly to show that o bg varatos the estmate of the stadard devato are to be epected. I cocluso: Based o the data from the aerobc test group we may epect that oe ca make measuremets of mamum torque (sometrc test), peak torque ad mamum agular velocty (dyamc test) wth HUR Leg eteso/curl such that the varato s less tha about 0 % 95 percet of the cases. Ths s e.g. o the same level as has bee reported for measuremets wth soketc maches. Tests wth other groups of subjects wll most lkely show the same patter f the measuremets are doe properly 9. No bg chage the rage of varato s to be epected, but oly ew data wll tell more about ths. For stace, subjects wth poor motor cotrol may be epected to show larger varatos the results. Mathematcal Apped Suppose the depedet radom varables (,,..., ) have the ormal dstrbuto N(, ) wth > > 0, the the probablty desty fucto (pdf) for (4) ˆ ˆ s gve qute accurately by 9 The test subjects should e.g. have eough tme to famlare themselves wth the equpmet before the measuremets. 6

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) (5) f ( ) + e 3 π + The followg pcture (Fgure 5) shows a smulated test seres based o geeratg a seres of 5 ( 5) ormally dstrbuted umbers ( 00, 5) ad calculatg the - umbers (). Ths s repeated 00 tmes gvg 500 datapots whose dstrbuto s plotted the fgure. The cotous le the pcture s obtaed usg 5, 00, ad 5 (ths correspods to the case 0.045) by umercally computg the dstrbuto of (5). Apparetly t fts the smulated dstrbuto very well. Fgure 5 Dstrbuto of smulated data. Cotous le s the dstrbuto computed from (5). For small (that s, for << ) (5) approaches a ormal dstrbuto wth the stadard devato gve by (6) It follows that (5) s well appromated by a ormal dstrbuto f we have 0 0 The codto says that the mea value must be cosderable larger tha the stadard devato of the average value. Ths s also a obvous requremet for (4) to be a meagful descrpto of varato. 7

HUR Techcal Report 000--9 verso.05 / Frak Borg (borgbros@ett.f) 8 (7) >> I the preset case (wth about 0.05 ad 5) we get for the left had sde (7) a umber aroud 000 so ths codto s well satsfed. Ths supports the use of (4) as a measure of varato ad the use of the ormal dstrbuto characterg the dstrbuto of the varato. A curous property of the dstrbuto (5) s that t lacks secod momet;.e, the stadard devato s ot defed for t (the tegral dverges). Ths s of o practcal cosequece here because very bg -values ( the tal of (5)) are physcally mpossble/rrelevat. Mathematcally oe may also adopt a alteratve defto of varato gve by (7) ˆ whch may be epected to be mathematcally better behaved certa respects. The varable (7) has geeral a complcated formula for the probabllty desty fucto. E.g. the case the varable (the square of that (7)) (8) ( ) y y + takes values the rage 0 to. If ad y are depedet radom varables wth the same N(,)-dstrbuto, the probablty desty fucto for (8) becomes (9) ( ) ( ) ( ) + 0 4!! 4 ) ( u u e du u e e f π π Ths dstrbuto s of terest oly for small values of /, cotrary to (5).