RHRC CONSORTIUM MONITORING AND EVALUATION TOOLKIT SURVEY SAMPLING TECHNIQUE EXAMPLE Instructions for Probability Proportional to Size

Sampling Technique
Prepared by Therese McGinn Hei br!nn Depar"#en" $% P$p! a"i$n and &a#i y Hea "h Mai #an Sch$$ $% P!b ic Hea "h' C$ !#bia Uni(ersi"y Probability proportional to size (PPS) is a sampling technique for use with surveys or minisurveys in which the probability of selecting a sampling unit (e.g., village, zone, district, health center) is proportional to the size of its population. t gives a probability (i.e., random, representative) sample. t is most useful when the sampling units vary considerably in size because it assures that those in larger sites have the same probability of getting into the sample as those in smaller sites, and vice verse. !his method also facilitates planning for field wor" because a predetermined number of respondents is interviewed in each unit selected, and staff can be allocated accordingly.

Steps in Applying Prob bility Proportion l to Si!e S "pling

#. $ist all villages in the pro%ect area (&olumn ' in the following e(ample) and their

populations (&olumn )). *ou can use the total population or the population of the group you are sampling, for e(ample, married women aged #+-,, or men #+--.. n a pinch, the number of households can be used.
/. &alculate the running cumulative population (&olumn &). !he last number in this column is

the total population of the pro%ect area. n the sample, the total population is #0,-#1.
2. 3etermine the number of sites which will be visited and the total sample size desired. 4or

a mini-survey, from which only basic frequencies are desired, e(pect to visit #.-#+ villages for a total sample size of #..-/... 4or a full scale survey, the sample size will be determined by the level and type of analysis planned5 you should probably e(pect to visit #+-2. sites, although the final number will be determined by the comple(ity of the area and the purpose of the study. n this e(ample, we will visit #. sites to conduct a mini-survey with a desired sample size of #+. women aged #+-,,. !hus, #+ women will be interviewed in each of the #. sites selected.
,. 3ivide the total population of the pro%ect area (#0,-#1, the final figure in &olumn &) by #.,

the number of sites we decide to visit. !he result, #,0-/, is called "he Sa#p in) In"er(a *SI+,
+. &hoose a number between # and the S at random. !his is the Rand$# S"ar" (6S). n this

sample, the 6S is #2/#.

RHRC C$ns$r"i!# M$ni"$rin) and E(a !a"i$n T$$ Ki" PPS Sa#p in) Techni1!e Oc"$ber -../' Pa)e 0

-. &alculate the following series7 6S5 6S 8 S 5 6S 8 /S 5 6S 8 2S 5 6S 8 ,S 5 6S 8 +S 5 6S 8

-S 5 6S 8 0S 5 6S 8 9S 5 6S 8 1S . :(ample7 6S 8 /S is to be calculated as / times the sampling interval added to the random start. n this case, #2/# 8 /(#0-/) ; ,9,+.
0. :ach of these #. numbers corresponds to a site on the list of villages. !he villages

selected are those for which &olumn &, the cumulative population, contains the numbers in the series we calculated. 4or e(ample, the first number in the series, #,2/#, is contained in village 2, which holds numbers 099 to #,9#1. !he second number in the series (2,.92) is contained in village -, which holds numbers /,1,2 to 2,/1,. &ontinuing in this manner, the desired number of sites will be selected. n this e(ample, the selected villages are numbers 2, -, 1, ##, #+, #9, /#, //, /+ and /1 (&olumn 3).
9. 's planned, #+ interviews will be conducted in each of the #. villages selected. Selection

of respondents within the village should also be done randomly, preferably from a list of eligible names or a map of households. f these are not available, estimate the number of households in the village from the population figures, then divide that number by #+, the number of respondents desired. !his is the interval, n. Starting from a random household, count every n"h household and interview all eligible respondents in that household. 4or e(ample, in village 2 there are #,.2/ people. f other information suggests that an average of - people ma"e up a household, then we estimate that there are #0/ household in the village (#.2/<-;#0/). !o get #+ respondents, we need to sample every ## th household (#0/<#+;##.+).

Notes

t can happen that a very large village contains more than one of the series of numbers. n this case, that village counts as two sites and twice the allocated number of interviews should be conducted there. =nce a household has been selected into the sample, a eligible respondents in the household should be interviewed. 4or e(ample, if the survey see"s women aged #+-,, and one household contains a mother, ,/, and her daughter, #1, then b$"h should be interviewed. !he reason for this is that if only one were interviewed, then a woman living in a household with another woman would have a lower chance (only # in /) of getting into the sample than a woman living with no other eligibles, whose chance is # in #. !he former group would be systematically deselected from the sample. Such a situation could arise with polygamous women, mothers and daughters or live-in servants. Such systematic bias could affect the data. t is better to e(ceed the sample size than not to reach it. n the above case, #+ households may result in more than #+ interviews. 3epending on the frequency of the situation, you can leave the plan as is and end up with more interviews than you really need or you can compensate for the multiple cases by sampling fewer households.

RHRC C$ns$r"i!# M$ni"$rin) and E(a !a"i$n T$$ Ki" PPS Sa#p in) Techni1!e

Oc"$ber -../' Pa)e -

Example: Drawing a Sample Using Probability Proportional to Size Sampling Technique olumn !
#illage # / 2 , + 0 9 1 #. ## #/ #2 #, #+ ##0 #9 #1 /. /# // /2 /, /+ //0 /9 /1 Sampling nterval (S ) ; ; ; 6andom Start (6S) Series numbers ; #0-/ #2/# #2/# 2.92 ,9,+ --.0 92-1 6S8+S 6S8-S 6S80S 6S89S 6S81S #.#2# ##912 #2-++ #+,#0 #0#01

olumn "
#illage Population +,/ /,+ #.2/ 9-0 /+2+/ 92+ -,+ ,/0 2#/ #2,/ 21. -., ,-+ 910 ,02-+ 1-0 +22 /#+ #+1. ,/2 -,+ 9-0 ,/2 #10 +92-+ 0+&umulative population #0-#1

olumn
umulati$e Population +,/ 090 #9#1 /-9/1,/ 2/1, ,#/1 ,00, +/.# ++#2 -9++ 0/,+ 09,1 92#, 1/## 1-90 #..+/ ##.#1 ##++/ ##0-0 #22+0 #209. #,,/+ #+/1/ #+0#+ #+1#/ #-,19 #-9-2 #0-#1 < < >umber of sites #.

olumn D
Series %umbers& Selecte' Sites

#2/# 2.92 ,9,+ --.0

92-1 #.#2# ##912 #2-++ #+,#0

#0#01

6S 6S8S 6S8/S 6S82S 6S8,S

Selecte' sites are $illages () *) +) ,,) ,-) ,.) /,) //) /-) /+

RHRC C$ns$r"i!# M$ni"$rin) and E(a !a"i$n T$$ Ki" PPS Sa#p in) Techni1!e

Oc"$ber -../' Pa)e 2