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.

506 QUANTITATIVE DATA ANALYSIS<br />

structure of a questionnaire survey very carefully<br />

in order to assist data entry for computer reading<br />

and analysis; an inappropriate layout may obstruct<br />

data entry and subsequent analysis by computer.<br />

The planning of data analysis will need to consider:<br />

<br />

<br />

What needs to be done with the data when<br />

they have been collected – how will they be<br />

processed and analysed<br />

How will the results of the analysis be verified,<br />

cross-checked and validated<br />

Decisions will need to be taken with regard to<br />

the statistical tests that will be used in data<br />

analysis as this will affect the layout of research<br />

items (for example in a questionnaire), and the<br />

computer packages that are available for processing<br />

quantitative and qualitative data, e.g. SPSS and<br />

NUD.IST respectively.<br />

Reliability<br />

We need to know how reliable is our instrument<br />

for data collection. Reliability in quantitative<br />

analysis takes two main forms, both of which<br />

are measures of internal consistency: the split-half<br />

technique and the alpha coefficient. Both calculate<br />

acoefficientofreliabilitythatcanliebetween0<br />

and 1. The split-half reliability has been discussed<br />

in an earlier chapter. The formula given is:<br />

r =<br />

2r<br />

1 + r<br />

where r = the actual correlation between the<br />

halves of the instrument (this requires the<br />

instrument to be able to be divided into two<br />

matched halves in terms of content and difficulty).<br />

So, for example, if the correlation coefficient<br />

between the two halves is 0.85 then the formula<br />

would be worked out thus:<br />

r = 2(0.85)<br />

1 + 0.85 = 1.70<br />

1.85 = 0.919<br />

Hence the split-half reliability coefficient is 0.919,<br />

which is very high. SPSS automatically calculates<br />

split-half reliability at the click of a button.<br />

An alternative calculation of reliability as<br />

internal consistency can be found in Cronbach’s<br />

alpha, frequently referred to simply as the alpha<br />

coefficient of reliability. The Cronbach alpha<br />

provides a coefficient of inter-item correlations,<br />

that is, the correlation of each item with the sum of<br />

all the other items. This is a measure of the internal<br />

consistency among the items (not, for example, the<br />

people). It is the average correlation among all the<br />

items in question, and is used for multi-item scales.<br />

SPSS calculates Cronbach’s alpha at the click of a<br />

button; the formula for alpha is:<br />

nr ii<br />

alpha =<br />

1 + (n − 1)r ii<br />

where n = the number of items in the test or<br />

survey (e.g. questionnaire) and r ii = the average<br />

of all the inter-item correlations. Let us imagine<br />

that the number of items in the survey is ten, and<br />

that the average correlation is 0.738. The alpha<br />

correlation can be calculated thus:<br />

alpha =<br />

nr ii<br />

10(.738)<br />

=<br />

1 + (n − 1)r ii 1 + (10 − 1).738<br />

= 7.38<br />

7.64 = 0.97<br />

This yields an alpha coefficient of 0.97, which<br />

is very high. The alpha coefficients are set out in<br />

Table 1 of the Appendices of Statistical Tables. For<br />

the split-half coefficient and the alpha coefficient<br />

the following guidelines can be used:<br />

>0.90 very highly reliable<br />

0.80–0.90 highly reliable<br />

0.70–0.79 reliable<br />

0.60–0.69 marginally/minimally reliable<br />

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

Saved successfully!

Ooh no, something went wrong!