what is the difference between random and non-random samplig?
1. Random sampling
a. Individual units from population are selected at random
b. Every individual has equal chance of being selected
c. Selection is left to chance factors
d. Suitable for homogenous and identical universe
e. Impartial and economical
f. Can be done in:
i. Lottery method
ii. Tables of random sample
o Free from personal bias
o Every item has equal chance of being selected
o Universe gets fairly represented
o Simple method
o Does not guarantee proportionate representation of different items in the universe
o Does not give weight age to certain important items in universe
2. Non-random sampling:
a. All units of population do not have equal chance of being selected
b. Convenience and judgment of the investigator plays a vital role in sample selection
c. Also known as purposive or deliberate sampling
d. Scientifically suitable when some of the items in the sample are of special significance and ought to be included in the sample.
o Selection tuned to purpose of study
o Possibility of personal bias
o Reliability of results, doubtful
a.Individual units from population are selected at random
b.Every individual has equal chance of being selected
c.Selection is left to chance factors
d.Suitable for homogenous and identical universe
e.Impartial and economical
f.Can be done in:
ii.Tables of random sample
oFree from personal bias
oEvery item has equal chance of being selected
oUniverse gets fairly represented
oDoes not guarantee proportionate representation of different items in the universe
oDoes not give weight age to certain important items in universe
a.All units of population do not have equal chance of being selected
b.Convenience and judgment of the investigator plays a vital role in sample selection
c.Also known as purposive or deliberate sampling
d.Scientifically suitable when some of the items in the sample are of special significance and ought to be included in the sample.
oSelection tuned to purpose of study
oPossibility of personal bias
oReliability of results, doubtfu
1- Each and every intem of univer has equal chances
2- it is free from personal biosness
3- Redom sampling more apropreat and homojinious data no persional biosness
1- Under this method depends judgment of investigator
2- it have more and more personal bioasness
3- it is not accuresi in collection data
It is generally agreed that the ) referred method of sampling
is the random method. The reason is that the behavior of the samples
taken randomly is known (i.e. follows central limit theorem predic-
Few people doing survey work, however, use the random method
because of prohibitive costs. The first step in random selection
requires numbering each member of the population. We once estimated
that to number the adult population of Saigon (which is often usod in
JUSPAO surveys) would take 40 man-weeks. Even the Americans would be
unwilling to foot the tab for labor.
Because of these practical considerations most people making
surveys use a sampling method that involves taking every nth member.
The purists cringe at this pointing out that each member has an equal
chance of being selected only once: at the time of the selection of
the starting point of n. How this affects the results is not known.
The essen?.?> of the argument for random sampling can be stated:
"We know what happens when you use the random method, but we don't know
what happens when you use non-randjm methods." The purpose of this
study is to find out what happens when non-random sampling methods are
I SAMPLING RANDOMLY DISTRIBUTED POPULATIONS
To compare the several methods we created a Vietnamese hamlet
consisting of 72 households strung along a river bank. (We find
this more interesting than creating a vector consisting of 72 colls.)
The hamlet consisted of 50% Catholics and 50% Buddhists assigned
randomly. In one series of tests the 36 Catholics were assigned one
per household (called the "without replacement hamlets"). The
question here is the proportion of Catholic households in the hamlet.
In another test the Catholics were assigned without the constraint of
one per household (called vhe "with replacement hamlets"). The
question here is the proportion of Catholics in the hamlet.
Constructing the hamlets in these two forms provides an analogue
to the most common types of survey data. The without replacement
hamlets represent the case where a single member of a household is
queried about his opinions or when binary choice response! are being
recorded: "Do you have children of school age?" The with replacement
hamlets is designed to represent the multiple response case: "How
many of your children arc- going to school?"
Our task was to estimate the proportions of Catholics using
several common sampling methods to determine which method was best.
In this study "best" has the specific meaning of having the greatest
accuracy when inferring the population mean. Note, carefully, that
"best" is not defined as agreeing most closely with central limit
theorem predictions. As used here "best" implies that the distribution
of the sample means has the smallest variance. method would be one in which every sample mean was identical to the
population mean (perfect representation with zero variance).
For each test 1C hamlets of each type were constructed and
sampled by random selection, both with and without replacement, and
by several regular (every nth)inethods. The regular method is easier to
depict than to describe. (The "1" indicates the household was queried.)
Ones 1...1...1...1...1...1... etc.
Twos 11 11 11 11.. etc.
XX.' ?? ? ? ?Xl??***tlX?**a?tXX
XX?* ? ???XX?XX? ?????XX? ?? ???XX????? ?XX
Notice that when 25% samples are taken four unique samples can be taken
for each sampling pattern. Similar sampling patterns were constructed
for Threes, Sixes and Nines patterns. Sampling using the Threes pattern
is a common practice. It is used to reduce travel which usually con-
sumes more than 50% of the data collection time. Clusters of six and
nine are never used but were included to exaggerate any effects of the