12.01.2015 Views

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

148 VALIDITY AND RELIABILITY<br />

by assigning items one to ten to one half and items<br />

eleven to twenty to the second half, may assign all<br />

the even numbered items to one group and all the<br />

odd numbered items to another. This would move<br />

towards the two halves being matched in terms of<br />

content and cumulative degrees of difficulty.<br />

An alternative measure of reliability as internal<br />

consistency is the Cronbach alpha, frequently<br />

referred to as the alpha coefficient of reliability, or<br />

simply the alpha. The Cronbach alpha provides a<br />

coefficient of inter-item correlations, that is, the<br />

correlation of each item with the sum of all the<br />

other relevant items, and is useful for multi-item<br />

scales. This is a measure of the internal consistency<br />

among the items (not, for example, the people). We<br />

address the alpha coefficient and its calculation in<br />

Part Five.<br />

Reliability, thus construed, makes several<br />

assumptions, for example that instrumentation,<br />

data and findings should be controllable,<br />

predictable, consistent and replicable. This<br />

presupposes a particular style of research,<br />

typically within the positivist paradigm. Cooper<br />

and Schindler (2001: 218) suggest that, in<br />

this paradigm, reliability can be improved by<br />

minimizing any external sources of variation:<br />

standardizing and controlling the conditions under<br />

which the data collection and measurement take<br />

place; training the researchers in order to ensure<br />

consistency (inter-rater reliability); widening the<br />

number of items on a particular topic; excluding<br />

extreme responses from the data analysis (e.g.<br />

outliers, which can be done with SPSS).<br />

Reliability in qualitative research<br />

While we discuss reliability in qualitative research<br />

here, the suitability of the term for qualitative<br />

research is contested (e.g. Winter 2000; Stenbacka<br />

2001; Golafshani 2003). Lincoln and Guba (1985)<br />

prefer to replace ‘reliability’ with terms such as<br />

‘credibility’, ‘neutrality’, ‘confirmability’, ‘dependability’,<br />

‘consistency’, ‘applicability’, ‘trustworthiness’<br />

and ‘transferability’, in particular the notion<br />

of ‘dependability’.<br />

LeCompte and Preissle (1993: 332) suggest that<br />

the canons of reliability for quantitative research<br />

may be simply unworkable for qualitative research.<br />

Quantitative research assumes the possibility of<br />

replication; if the same methods are used with the<br />

same sample then the results should be the same.<br />

Typically quantitative methods require a degree<br />

of control and manipulation of phenomena. This<br />

distorts the natural occurrence of phenomena (see<br />

earlier: ecological validity). Indeed the premises of<br />

naturalistic studies include the uniqueness and<br />

idiosyncrasy of situations, such that the study<br />

cannot be replicated – that is their strength rather<br />

than their weakness.<br />

On the other hand, this is not to say that<br />

qualitative research need not strive for replication<br />

in generating, refining, comparing and validating<br />

constructs (see http://www.routledge.com/<br />

textbo<strong>ok</strong>s/9780415368780 – Chapter 6, file 6.7.<br />

ppt). Indeed LeCompte and Preissle (1993: 334)<br />

argue that such replication might include repeating<br />

<br />

<br />

<br />

<br />

<br />

the status position of the researcher<br />

the choice of informant/respondents<br />

the social situations and conditions<br />

the analytic constructs and premises that are<br />

used<br />

the methods of data collection and analysis.<br />

Further, Denzin and Lincoln (1994) suggest that<br />

reliability as replicability in qualitative research<br />

can be addressed in several ways:<br />

<br />

<br />

<br />

stability of observations: whether the researcher<br />

would have made the same observations and<br />

interpretation of these if they had been<br />

observed at a different time or in a different<br />

place<br />

parallel forms: whether the researcher would<br />

have made the same observations and<br />

interpretations of what had been seen if he<br />

or she had paid attention to other phenomena<br />

during the observation<br />

inter-rater reliability: whether another observer<br />

with the same theoretical framework and<br />

observing the same phenomena would have<br />

interpreted them in the same way.<br />

Clearly this is a contentious issue, for it is seeking<br />

to apply to qualitative research the canons of

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

Saved successfully!

Ooh no, something went wrong!