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.

MEASURES OF DIFFERENCE BETWEEN GROUPS AND MEANS 547<br />

Box 24.42<br />

The paired samples t-test<br />

Paired samples test<br />

Paired differences<br />

95 % confidence<br />

interval of the<br />

difference<br />

Chapter 24<br />

Pair 1<br />

The attention given to<br />

teaching and learning<br />

at the school – the<br />

quality of the lesson<br />

preparation<br />

Mean SD SE mean Lower Upper t df Sig. (2-tailed)<br />

−1.69 2.430 0.077 −1.84 −1.54 −21.936 999 0.000<br />

governors) and one is a continuous variable (e.g.<br />

marks on a test).<br />

One-way analysis of variance<br />

Let us imagine that we have four types of school:<br />

rural primary, rural secondary, urban primary and<br />

urban secondary. Let us imagine further that all<br />

of the schools in these categories have taken<br />

the same standardized test of mathematics, and<br />

the results have been given as a percentage thus<br />

(Box 24.43).<br />

The table gives us the means, standard<br />

deviations, standard error, confidence intervals,<br />

and the minimum and maximum marks for each<br />

group. At this stage we are interested only in the<br />

means:<br />

Rural primary:<br />

Rural secondary:<br />

Urban primary:<br />

Urban secondary:<br />

mean = 59.85 per cent<br />

mean = 60.44 per cent<br />

mean = 50.64 per cent<br />

mean = 51.70 per cent<br />

Are these means statistically significantly different<br />

Analysis of variance will tell us whether<br />

they are. We commence with the null hypothesis<br />

(‘there is no statistically significant difference<br />

between the four means’) and then we set the<br />

level of significance (α) to use for supporting or<br />

not supporting the null hypothesis; for example<br />

we could say ‘Let α = 0.05’. SPSS calculates the<br />

following (Box 24.44).<br />

This tells us that, for three degrees of freedom<br />

(df), the F-ratio is 8.976. The F-ratio is the between<br />

group mean square (variance) divided by the within<br />

group mean square (variance), i.e.:<br />

F =<br />

Between group variance<br />

Within group variance = 3981.040<br />

443.514 = 8.976<br />

By lo<strong>ok</strong>ing at the final column (‘Sig.’) ANOVA<br />

tell us that there is a statistically significant<br />

difference between the means (ρ = 0.000). This<br />

does not mean that all the means are statistically<br />

significantly different from each other, but that<br />

some are. For example, it may be that the means<br />

for the rural primary and rural secondary schools<br />

(59.85 per cent and 60.44 per cent respectively)<br />

are not statistically significantly different, and that<br />

the means for the urban primary schools and urban<br />

secondary schools (50.64 per cent and 51.70 per<br />

cent respectively) are not statistically significantly<br />

different. However, it could be that there is<br />

astatisticallysignificantdifferencebetweenthe<br />

scores of the rural (primary and secondary) and<br />

the urban (primary and secondary) schools. How<br />

can we find out which groups are different from<br />

each other<br />

There are several tests that can be employed<br />

here, though we will only concern ourselves with<br />

averycommonlyusedtest:theTukeyhonestly

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

Saved successfully!

Ooh no, something went wrong!