R help - Mar 2009 - (Interpretation) VGAM - FRECHET 3 parameters by maximum likelihood estimation for

Dear R Helpers


This is the R code (which I have slightly changed) I got in VGAM package for
estimating the parameters of FRECHET.

_________________________________________________________________

y = rfrechet(n <- 100, shape=exp(exp(0)))       #    (A)

fit3 = vglm(y ~ 1, frechet3(ilocation=0), trace=TRUE, maxit=155)   #   (B)

coef(fit3, matrix=TRUE)                               #   (C)

Coef(fit3)                                            #   (D)

fitted(fit3)[1:5,]                                    #   (E)

mean(y)                                               #   (F) 

weights(fit3, type="w")[1:5,]                         #   (G) 
                             

vcov(fit3)                                            #    (H)    

fit3 at extra$location[1:3]  # Estimate of the location parameter  #    (I)
    
fit3 at extra$LHSanchor      # Anchor Point              #    (J)

_________________________________________________________________

When I run these commands using R I get following output corresponding to each
command line


OUTPUT # (A)

[1]  0.7495887  1.1578511  0.8872487  0.7781103  1.1026540  1.8557026  1.1527270
1.1538885

[9]  1.0436005  0.8702552  1.0326653  1.1456378  1.1292287  0.9779141  0.7882611
2.0040318

[17]  0.8832376  1.6684632  0.7112340  1.8440806  0.8103669  0.6171517 
0.6632612  0.9288173

[25]  1.2625774  1.2371949  2.3995687  0.9674406  1.2205647  2.2997349 
1.3561988  1.0257294

[33]  0.7693035  1.3112102  1.2841138  0.5207330  0.8861406  0.7819666 
1.1376008  1.3396068

[41]  0.6828540  0.7358596  0.8678724  1.1191930  0.9581007  0.8645687 
1.0269345  0..9853376

[49]  0.7861684  2.2938797  0.5994656  1.3733527  3.0904095  2.7179989 
0.8408945  0.9995262

[57]  1.0456575  0.6139744 17.4274447  0.5928158  1.6172649  0.7354806 
1.4714551  0.7851852

[65]  1.9091740  1.0862915  1.9082128  1.0051395  1.0991948  1.6698645 
2.8901334  0.8174091

[73]  1.3964180  0.7452934  0.8133906  1.2417527  2.1583672  1.1822409 
0.8068691  0.8410480

[81]  1.2394454  1.3391757  1.0867579  1.3395385  1.2791401  0.7897431 
1.2112373  1.6531473

[89]  1.7432403  6.7756058  0.9385585  1.5561437  0.9590773  0.7936483 
2.5215168  1.1984615

[97]  0.6345925  1.4218795  2.3178357  1.5454827



OUTPUT # (B)

In vglm.fitter(x = x, y = y, w = w, offset = offset, Xm2 = Xm2,   :  convergence
not obtained in 155 iterations.




OUTPUT # (C) 
> coef(fit3, matrix=TRUE)                               
                log(difference)	 log(scale)	 loglog(shape)
(Intercept)      -0.7954072	 -0.1283776  	 -0.03175329




OUTPUT # (D)
> Coef(fit3))
                 difference 	scale      	shape 
                  0.4513974  	0.8795212  	2.6346374

(IS IT THAT THESE GIVE THE SCALE AND SHAPE PARAMETER  ESTIMATES OF FRECHET.?)



OUTPUT # (E)
> fitted(fit3)[1:5,]          
[1] 1.339454  1.339454  1.33954  1.339454  1.339454


(WHAT DOES THIS MEAN?)



OUTPUT # (F)
> mean(y) 
[1] 1.439024



OUTPUT # (G) 	

weights(fit3, type="w")[1:5,]
         

 [,1]    		 [,2]      		[,3] 		[,4]      		[,5]       		[,6]

[1,] 24.264915 	46.60547 	38.864660 	-33.624853 	40.492033 	-29.075460

[2,]  7.599936 	14.18608  	1.554636  	-9.724618  	2.440009  	-2.670739

[3,]  6.486803 	26.53437  	1.855081 	-12.926846 	-5.700736   	2.433455

[4,] 16.768849 	37.36856 	36.299447 	-25.025068 	34.078418 	-22.643518

[5,]  9.562077 	15.87083  	2.671884 	-11.625172  	4.418292  	-4.454716



(WHAT DOES THIS MEAN?)



OUTPUT # (H)

vcov(fit3)                                                      
                   (Intercept)1  (Intercept)2	 (Intercept):3
(Intercept) : 1	  0.0022605491   0.0021317673 	 -1.576056e-04
(Intercept) : 2	  0.0021317673   0.0020926942    -1.780387e-04
(Intercept) : 3	 -0.0001576056	-0.0001780387     3.892675e-05


Warning message:

In summaryvlm(object, corr = FALSE, dispersion = dispersion) :   the estimated
variance-covariance matrix is usually inaccurate as the working weight matrices
are a crude BFGS quasi-Newton approximation


OUTPUT # (I)	

fit3 at extra$location[1:3] # Estimate of the location parameter 

         1         	 2          		3 
0.06930899	 0.06930899 	0.06930899

(What does this MEAN? IS IT THAT 'D' gives SCALE and SHAPE parameter
whereas this gives me LOCATION parameter?)



OUTPUT # (J)	

fit3 at extra$LHSanchor # Anchor point                                 

[1] 0.520733

(What does this mean?)



My problem is to calculate the estimators of 3 Parameter FRECHET distribution.
Which of the above figures give me estimates of the three parameters of the
FRECHET distribution.



With regards and thanking in advance,


Maithili




      Connect with friends all over the world. Get Yahoo!