SlideShare a Scribd company logo
1 of 10
Download to read offline
Theorem 1: Invariance property 
of maximum-likelihood 
estimators 
Benjie Lloyd V. Tabios 
September 11, 2014
1 
Invariance property of maximum-likelihood 
estimators 
Let ^ 
= ^#(X1;X2; : : : ;Xn) be the maximum- 
likelihood estimator of  in the density f(x; ), 
where  is assumed unidimensional. If (:) is a 
function with a single-valued inverse, then the 
maximum-likelihood estimator of () is ( ^ ). 
1
2 
For example, in the normal density with 110 
known the maximum-likelihood estimator of 2 
is 
(1) 1=n 
(Xi  o)2 
Xn 
i=1 
By the invariance property of maximum-likelihood 
estimators, the maximum- likelihood estimator 
of  is 
vuut 
(2) 1=n 
(Xi  o)2 
Xn 
i=1 
Similarly, the maximum-likelihood estimator of, 
say, log 2 is 
log 
h 
1=n 
Xn 
i=1 
(Xi  o)2 
i 
(3) 
2
3 
The invariance property of maximum-likelihood 
estimators that is exhibited in Theorem 1 above 
can and should be extended. Following Zehna 
[43], we extend in two directions: First.  
will be taken as k-dimensional rather than uni- 
dimensional, and, second, the assumption that 
(:) has a single-valued inverse will be removed. 
3
4 
It can be noted that such extension is neces- 
sary by considering two simple examples. As a
rst example, suppose an estimate of the vari- 
ance, namely (1  ), of a Bernoulli distribu- 
tion is desired. Example 5 gives the maximum- 
likelihood estimate of  to be x, but since 
(1) is not a one-to;'one function of , The- 
orem 1 does not give the maximum-likelihood 
estimator of (1  ). 
4
5 
Theorem 2 will give such an estimate, and 
it will be x(1  x). As a second example, 
consider sampling from a normal distribution 
where both  and 2 are unknown, and sup- 
pose an estimate of [X2] = 2+2 is desired. 
Example 6 gives the maximum-likelihood esti- 
mates of  and 2, but 2+2 is not a one-to- 
one function of  and 2, and so the maximum- 
likelihood estimate of 2 + 2 is not known. 
Such an estimate will be obtainable from The- 
orem 2. It will be x2 +(1=n) 
P 
(xi  x)2. 
5
6 
Let  = (1; : : : ; k) be a k-dimensional parame- 
ter, and, as before, let  
denote the parameter 
space. Suppose that the maximum-likelihood 
estimate of () = (1(); : : : ; r()), where 1  
r  k, is desired. Let T denote the range space 
of the transformation () = (1(); : : : ; r()). 
T is an r-dimensional space. De
ne 
M(; x1; : : : ; xn) = supf:()=g L(; x1; : : : ; xn). 
M(; x1; : : : ; xn)is called the likelihood function 
induced by ().* 
6
7 
When estimating  we maximized the likeli- 
hood function L(; x1; : : : ; xn) as a function of 
 for

More Related Content

What's hot

Newton raphson method
Newton raphson methodNewton raphson method
Newton raphson methodMeet Patel
 
05b distribution functions
05b distribution functions05b distribution functions
05b distribution functionsnumanakhtar2
 
Newton raphson method
Newton raphson methodNewton raphson method
Newton raphson methodJayesh Ranjan
 
Q1 create a java desktop application to find the largest number among the t...
Q1  create a java desktop application to find the largest number  among the t...Q1  create a java desktop application to find the largest number  among the t...
Q1 create a java desktop application to find the largest number among the t...Manoj Bhakuni
 
Newton raphsonmethod presentation
Newton raphsonmethod presentationNewton raphsonmethod presentation
Newton raphsonmethod presentationAbdullah Moin
 
Fast lookup in sorted array jakob voigts
Fast lookup in sorted array   jakob voigtsFast lookup in sorted array   jakob voigts
Fast lookup in sorted array jakob voigtsKyle Cheng
 
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...Establishing the Equation of the Wave Function and the System Model (ρ, m, an...
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...BRNSS Publication Hub
 
algebric solutions by newton raphson method and secant method
algebric solutions by newton raphson method and secant methodalgebric solutions by newton raphson method and secant method
algebric solutions by newton raphson method and secant methodNagma Modi
 
Teachingtools4kidsbasic/calculus
Teachingtools4kidsbasic/calculusTeachingtools4kidsbasic/calculus
Teachingtools4kidsbasic/calculusOpenCourse
 

What's hot (19)

Newton raphson method
Newton raphson methodNewton raphson method
Newton raphson method
 
05b distribution functions
05b distribution functions05b distribution functions
05b distribution functions
 
Lec8
Lec8Lec8
Lec8
 
Newton
NewtonNewton
Newton
 
Newton raphson method
Newton raphson methodNewton raphson method
Newton raphson method
 
Q1 create a java desktop application to find the largest number among the t...
Q1  create a java desktop application to find the largest number  among the t...Q1  create a java desktop application to find the largest number  among the t...
Q1 create a java desktop application to find the largest number among the t...
 
Newton raphsonmethod presentation
Newton raphsonmethod presentationNewton raphsonmethod presentation
Newton raphsonmethod presentation
 
Computer Network Assignment Help
Computer Network Assignment HelpComputer Network Assignment Help
Computer Network Assignment Help
 
Fast lookup in sorted array jakob voigts
Fast lookup in sorted array   jakob voigtsFast lookup in sorted array   jakob voigts
Fast lookup in sorted array jakob voigts
 
Lecture1 3
Lecture1 3Lecture1 3
Lecture1 3
 
Soft Heaps
Soft HeapsSoft Heaps
Soft Heaps
 
Newton Raphson
Newton RaphsonNewton Raphson
Newton Raphson
 
1508 calculus-fundamental theorem
1508 calculus-fundamental theorem1508 calculus-fundamental theorem
1508 calculus-fundamental theorem
 
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...Establishing the Equation of the Wave Function and the System Model (ρ, m, an...
Establishing the Equation of the Wave Function and the System Model (ρ, m, an...
 
Code
CodeCode
Code
 
algebric solutions by newton raphson method and secant method
algebric solutions by newton raphson method and secant methodalgebric solutions by newton raphson method and secant method
algebric solutions by newton raphson method and secant method
 
Prml 2.3
Prml 2.3Prml 2.3
Prml 2.3
 
Teachingtools4kidsbasic/calculus
Teachingtools4kidsbasic/calculusTeachingtools4kidsbasic/calculus
Teachingtools4kidsbasic/calculus
 
Taylor series
Taylor seriesTaylor series
Taylor series
 

Viewers also liked (11)

Magazine inspiration
Magazine inspirationMagazine inspiration
Magazine inspiration
 
Flat plan
Flat planFlat plan
Flat plan
 
hotel design research
hotel design researchhotel design research
hotel design research
 
two
twotwo
two
 
Untitled Presentation
Untitled PresentationUntitled Presentation
Untitled Presentation
 
three
threethree
three
 
lloydtabios2014
lloydtabios2014lloydtabios2014
lloydtabios2014
 
Room Style
Room StyleRoom Style
Room Style
 
Nnnnn
NnnnnNnnnn
Nnnnn
 
Audience profile
Audience profileAudience profile
Audience profile
 
Those winter sundays by robert hayden
Those winter sundays by robert haydenThose winter sundays by robert hayden
Those winter sundays by robert hayden
 

Similar to benjielloyd1234

Moment-Generating Functions and Reproductive Properties of Distributions
Moment-Generating Functions and Reproductive Properties of DistributionsMoment-Generating Functions and Reproductive Properties of Distributions
Moment-Generating Functions and Reproductive Properties of DistributionsIJSRED
 
IJSRED-V2I5P56
IJSRED-V2I5P56IJSRED-V2I5P56
IJSRED-V2I5P56IJSRED
 
this materials is useful for the students who studying masters level in elect...
this materials is useful for the students who studying masters level in elect...this materials is useful for the students who studying masters level in elect...
this materials is useful for the students who studying masters level in elect...BhojRajAdhikari5
 
Random variables
Random variablesRandom variables
Random variablesMenglinLiu1
 
Probability distribution
Probability distributionProbability distribution
Probability distributionManoj Bhambu
 
Limits And Derivative
Limits And DerivativeLimits And Derivative
Limits And DerivativeAshams kurian
 
Limits And Derivative slayerix
Limits And Derivative slayerixLimits And Derivative slayerix
Limits And Derivative slayerixAshams kurian
 
Expectation of Discrete Random Variable.ppt
Expectation of Discrete Random Variable.pptExpectation of Discrete Random Variable.ppt
Expectation of Discrete Random Variable.pptAlyasarJabbarli
 
Probability and Statistics
Probability and StatisticsProbability and Statistics
Probability and StatisticsMalik Sb
 
Qt random variables notes
Qt random variables notesQt random variables notes
Qt random variables notesRohan Bhatkar
 
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...Christian Robert
 
Communication Theory - Random Process.pdf
Communication Theory - Random Process.pdfCommunication Theory - Random Process.pdf
Communication Theory - Random Process.pdfRajaSekaran923497
 
Interpolation techniques - Background and implementation
Interpolation techniques - Background and implementationInterpolation techniques - Background and implementation
Interpolation techniques - Background and implementationQuasar Chunawala
 
Quantitative Techniques random variables
Quantitative Techniques random variablesQuantitative Techniques random variables
Quantitative Techniques random variablesRohan Bhatkar
 

Similar to benjielloyd1234 (20)

Moment-Generating Functions and Reproductive Properties of Distributions
Moment-Generating Functions and Reproductive Properties of DistributionsMoment-Generating Functions and Reproductive Properties of Distributions
Moment-Generating Functions and Reproductive Properties of Distributions
 
IJSRED-V2I5P56
IJSRED-V2I5P56IJSRED-V2I5P56
IJSRED-V2I5P56
 
Si report
Si reportSi report
Si report
 
this materials is useful for the students who studying masters level in elect...
this materials is useful for the students who studying masters level in elect...this materials is useful for the students who studying masters level in elect...
this materials is useful for the students who studying masters level in elect...
 
Random variables
Random variablesRandom variables
Random variables
 
Probability distribution
Probability distributionProbability distribution
Probability distribution
 
Ch5
Ch5Ch5
Ch5
 
Limits And Derivative
Limits And DerivativeLimits And Derivative
Limits And Derivative
 
Limits And Derivative slayerix
Limits And Derivative slayerixLimits And Derivative slayerix
Limits And Derivative slayerix
 
Expectation of Discrete Random Variable.ppt
Expectation of Discrete Random Variable.pptExpectation of Discrete Random Variable.ppt
Expectation of Discrete Random Variable.ppt
 
Probability and Statistics
Probability and StatisticsProbability and Statistics
Probability and Statistics
 
Qt random variables notes
Qt random variables notesQt random variables notes
Qt random variables notes
 
Unit 1
Unit 1Unit 1
Unit 1
 
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...
Statistics (1): estimation Chapter 3: likelihood function and likelihood esti...
 
Communication Theory - Random Process.pdf
Communication Theory - Random Process.pdfCommunication Theory - Random Process.pdf
Communication Theory - Random Process.pdf
 
Interpolation techniques - Background and implementation
Interpolation techniques - Background and implementationInterpolation techniques - Background and implementation
Interpolation techniques - Background and implementation
 
CH6.pdf
CH6.pdfCH6.pdf
CH6.pdf
 
Ch6
Ch6Ch6
Ch6
 
Quantitative Techniques random variables
Quantitative Techniques random variablesQuantitative Techniques random variables
Quantitative Techniques random variables
 
Exponential function
Exponential functionExponential function
Exponential function
 

benjielloyd1234

  • 1. Theorem 1: Invariance property of maximum-likelihood estimators Benjie Lloyd V. Tabios September 11, 2014
  • 2. 1 Invariance property of maximum-likelihood estimators Let ^ = ^#(X1;X2; : : : ;Xn) be the maximum- likelihood estimator of in the density f(x; ), where is assumed unidimensional. If (:) is a function with a single-valued inverse, then the maximum-likelihood estimator of () is ( ^ ). 1
  • 3. 2 For example, in the normal density with 110 known the maximum-likelihood estimator of 2 is (1) 1=n (Xi o)2 Xn i=1 By the invariance property of maximum-likelihood estimators, the maximum- likelihood estimator of is vuut (2) 1=n (Xi o)2 Xn i=1 Similarly, the maximum-likelihood estimator of, say, log 2 is log h 1=n Xn i=1 (Xi o)2 i (3) 2
  • 4. 3 The invariance property of maximum-likelihood estimators that is exhibited in Theorem 1 above can and should be extended. Following Zehna [43], we extend in two directions: First. will be taken as k-dimensional rather than uni- dimensional, and, second, the assumption that (:) has a single-valued inverse will be removed. 3
  • 5. 4 It can be noted that such extension is neces- sary by considering two simple examples. As a
  • 6. rst example, suppose an estimate of the vari- ance, namely (1 ), of a Bernoulli distribu- tion is desired. Example 5 gives the maximum- likelihood estimate of to be x, but since (1) is not a one-to;'one function of , The- orem 1 does not give the maximum-likelihood estimator of (1 ). 4
  • 7. 5 Theorem 2 will give such an estimate, and it will be x(1 x). As a second example, consider sampling from a normal distribution where both and 2 are unknown, and sup- pose an estimate of [X2] = 2+2 is desired. Example 6 gives the maximum-likelihood esti- mates of and 2, but 2+2 is not a one-to- one function of and 2, and so the maximum- likelihood estimate of 2 + 2 is not known. Such an estimate will be obtainable from The- orem 2. It will be x2 +(1=n) P (xi x)2. 5
  • 8. 6 Let = (1; : : : ; k) be a k-dimensional parame- ter, and, as before, let denote the parameter space. Suppose that the maximum-likelihood estimate of () = (1(); : : : ; r()), where 1 r k, is desired. Let T denote the range space of the transformation () = (1(); : : : ; r()). T is an r-dimensional space. De
  • 9. ne M(; x1; : : : ; xn) = supf:()=g L(; x1; : : : ; xn). M(; x1; : : : ; xn)is called the likelihood function induced by ().* 6
  • 10. 7 When estimating we maximized the likeli- hood function L(; x1; : : : ; xn) as a function of for
  • 11. xed x1; : : : ; xn; when estimating = () we will maximize the likelihood function in- duced by (), namely M(; x1; : : : ; xn) as a function of for
  • 12. xed x1; : : : ; xn. Thus, the maximum-likelihood estimate of = (), de- noted by ^, is any value that maximizes the induced likelihood function for
  • 13. xed x1; : : : ; xn; that is, ^ is such that M(^; x1; : : : ; xn) M(; x1; : : : ; xn) for all 2 T. The invariance property of maximum- likelihood estimation is given in the following theorem. 7
  • 14. 8 in addition...... The maximum likelihood estimate is invariant under functional transformations. That is, if T = t(X1; : : : ;Xn) is the mle of and if u() is a function of , then u(T) is the mle of u(). For example, if ^ is the mle of , then ^2 is the mle of 2. That is, c2 = ^2. 8
  • 15. 9 THANK YOU and GOD BLESS !!!! :) 9