12.01.2015 Views

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

SAMPLING ERROR 107<br />

Box 4.2<br />

Distribution of sample means showing the spread<br />

of a selection of sample means around the<br />

population mean<br />

Mpop Population mean<br />

Ms Ms Ms Ms Mpop Ms Ms Ms Ms<br />

Source: Cohen and Holliday 1979<br />

Ms<br />

Sample means<br />

Chapter 10). Rose and Sullivan (1993: 144)<br />

remind us that 95 per cent of all sample means<br />

fall between plus or minus 1.96 standard errors<br />

of the sample and population means, i.e. that we<br />

have a 95 per cent chance of having a single<br />

sample mean within these limits, that the sample<br />

mean will fall within the limits of the population<br />

mean.<br />

By drawing a large number of samples of equal<br />

size from a population, we create a sampling<br />

distribution. We can calculate the error involved<br />

in such sampling (see http://www.routledge.<br />

com/textbo<strong>ok</strong>s/9780415368780 – Chapter 4, file<br />

4.5.ppt). The standard deviation of the theoretical<br />

distribution of sample means is a measure of<br />

sampling error and is called the standard error<br />

of the mean (SE M ). Thus,<br />

SE = SD s<br />

√<br />

N<br />

where SD S = the standard deviation of the sample<br />

and N = the number in the sample.<br />

Strictly speaking, the formula for the standard<br />

error of the mean is:<br />

SE = SD pop<br />

√<br />

N<br />

where SD pop = the standard deviation of the<br />

population.<br />

However, as we are usually unable to ascertain the<br />

SD of the total population, the standard deviation<br />

of the sample is used instead. The standard error<br />

of the mean provides the best estimate of the<br />

sampling error. Clearly, the sampling error depends<br />

on the variability (i.e. the heterogeneity) in the<br />

population as measured by SD pop as well as the<br />

sample size (N) (RoseandSullivan1993:143).<br />

The smaller the SD pop the smaller the sampling<br />

error; the larger the N, thesmallerthesampling<br />

error. Where the SD pop is very large, then N<br />

needs to be very large to counteract it. Where<br />

SD pop is very small, then N, too,canbesmall<br />

and still give a reasonably small sampling error.<br />

As the sample size increases the sampling error<br />

decreases. Hopkins et al. (1996:159)suggestthat,<br />

unless there are some very unusual distributions,<br />

samples of twenty-five or greater usually yield a<br />

normal sampling distribution of the mean. For<br />

further analysis of steps that can be taken to cope<br />

with the estimation of sampling in surveys we refer<br />

the reader to Ross and Wilson (1997).<br />

The standard error of proportions<br />

We said earlier that one answer to ‘How big a<br />

sample must I obtain’ is ‘How accurate do I want<br />

my results to be’ This is well illustrated in the<br />

following example:<br />

Aschoolprincipalfindsthatthe25studentsshetalks<br />

to at random are reasonably in favour of a proposed<br />

change in the lunch break hours, 66 per cent being in<br />

favour and 34 per cent being against. How can she be<br />

sure that these proportions are truly representative of<br />

the whole school of 1,000 students<br />

Asimplecalculationofthestandarderrorof<br />

proportions provides the principal with her answer.<br />

√<br />

P × Q<br />

SE =<br />

N<br />

where<br />

P = the percentage in favour<br />

Q = 100 per cent − P<br />

N = the sample size<br />

Chapter 4

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!