Sampling

can not measure

can be measured

e.g. number of pets

finite

e.g.length of rods

collection of items we are interested in

list of members in population

individual member within frame

measure of entire population

collection of individuals or items from a population

each item has equal chance of selection.

only method free of bias

simple

quick

easy to use

cheap

not suitable for large samples

randomly select a number k which is less than pop size.

then choose every k’th number

simple to use

suitable for large sample

could be bias

population divided into natural occuring mututally exclusive groups whose members have something in common.

random sample chosen from each strata in proportion to size of strata

pros:

representative of population structure

seperate mean and variance estimates for strata

cons:

difficult to obtain sampling frame

interview certain no. of ppl with characteristics. Once quota is filled ignore rest of population

pros:

minimum costs

no frame or list

cons:

not random

interviewer bias

people who are available at time of study

pros:

easy to carry out

cons:

dependant on individual researcher

1. allocate 2 digit number to everyone 00 - 99

2. select random starting point

3. select 12 numbers

4. if repeat then ignore

5. go back to original population and select corresponding ppl

total = 200

men = ( 120 / 200 ) x 40 = 24

women = ( 80 / 200 ) x 40 = 16

1. label men 000 to 119

2. label women 00 to 79

3. random number table or balls to select 24 men

4. random number table or balls to select 16 women