The last example illustrates a situation in which two units must be assigned 

 to the stratum for measurement with probability equal to one. The remaining three 

 sample units happen to include all of the remaining "large" units in the population. 

 Nevertheless, the smaller units could have been a part of the sample had the random 

 number been small enough. 



ID 



3 

 4 

 5 

 6 

 7 

 8 

 9 

 10 

 11 

 1 

 2 



EST. SIZF 



3.00000 

 3. OOCOO 

 5. OOCOO 

 5. OCOOO 

 109.00000 

 109. 00000 

 115. OOCOO 

 13C. 00000 

 999. 00000 

 1. 00000 

 1. 5GGC0 



NO. OF UNITS TO BE 

 MEASURED = 5 



SAMPLES DRAWN 



( NO. I N STRATUM = 9 ) 



{ NO. CF SAMPLES = 3) 



( TX = 0.3515000E 03) 

 ( TS = 0. 1206964E-01 ) 



( TP = 0.2865710E 00) 



ID PROBABILITY 



1 I 



10 

 7 



8 

 9 



0. 1 OOOOOOE 

 O.IOOOOOOE 

 0.9302982E 

 0.9302988E 

 0.98 15083E 



01 

 01 



00 

 00 

 00 



LITERATURE CITED 



Grosenbaugh, L. R. 



1969. More on FORTRAN random number generators. Commun . ACM 12(11): 639. 

 Hartley, H. 0. 



1966. Systematic sampling with unequal probability and without replacement. 

 J. Amer. Statist. Assoc. 61: 739-748. 



Marsaglia, George, and T. A. Bray. 



1968. One- line random number generators and their use in combinations. Commun. 

 ACM 11(11) : 757-759. 



12 



