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.

276 EXPERIMENTS AND META-ANALYSIS<br />

and, hence, inferring causality is more contestable,<br />

but they have the attraction of taking place in a<br />

natural setting. Extraneous variables may include,<br />

for example:<br />

<br />

<br />

<br />

participant factors: they may differ on important<br />

characteristics between the control and<br />

experimental groups<br />

intervention factors: the intervention may not<br />

be exactly the same for all participants, varying,<br />

for example, in sequence, duration, degree<br />

of intervention and assistance, and other<br />

practices and contents<br />

situational factors: the experimental conditions<br />

may differ.<br />

These can lead to experimental error, in which<br />

the results may not be due to the independent<br />

variables in question.<br />

The pretest-post-test control and<br />

experimental group design<br />

A complete exposition of experimental designs<br />

is beyond the scope of this chapter. In<br />

the brief outline that follows, we have selected<br />

one design from the comprehensive treatment<br />

of the subject by Campbell and Stanley<br />

(1963) in order to identify the essential features<br />

of what they term a ‘true experimental’<br />

and what Kerlinger (1970) refers to as a ‘good’<br />

design. Along with its variants, the chosen design<br />

is commonly used in educational experimentation<br />

(see http://www.routledge.com/textbo<strong>ok</strong>s/<br />

9780415368780 – Chapter 13, file 13.3. ppt).<br />

The pretest-post-test control group design can<br />

be represented as:<br />

Experimental RO 1 X O 2<br />

Control RO 3 O 4<br />

Kerlinger (1970) observes that, in theory, random<br />

assignment to E and C conditions controls all<br />

possible independent variables. In practice, of<br />

course, it is only when enough subjects are<br />

included in the experiment that the principle<br />

of randomization has a chance to operate as<br />

a powerful control. However, the effects of<br />

Box 13.2<br />

The effects of randomization<br />

Select twenty cards from a pack, ten red and ten black.<br />

Shuffle and deal into two ten-card piles. Now count the<br />

number of red cards and black cards in either pile and<br />

record the results. Repeat the whole sequence many<br />

times, recording the results each time.<br />

You will soon convince yourself that the most likely<br />

distribution of reds and blacks in a pile is five in each:<br />

the next most likely, six red (or black) and four black<br />

(or red); and so on. You will be lucky (or unlucky for the<br />

purposes of the demonstration!) to achieve one pile of<br />

red and the other entirely of black cards. The probability<br />

of this happening is 1 in 92,378. On the other hand, the<br />

probability of obtaining a ‘mix’ of not more than six of<br />

one colour and four of the other is about 82 in 100.<br />

If you now imagine the red cards to stand for the<br />

‘better’ ten children and the black cards for the ‘poorer’<br />

ten children in a class of twenty, you will conclude that<br />

the operation of the laws of chance alone will almost<br />

probably give you close equivalent ‘mixes’ of ‘better’<br />

and ‘poorer’ children in the experimental and control<br />

groups.<br />

Source:adaptedfromPilliner1973<br />

randomization even with a small number of<br />

subjects is well illustrated in Box 13.2.<br />

Randomization, then, ensures the greater<br />

likelihood of equivalence, that is, the apportioning<br />

out between the experimental and control groups<br />

of any other factors or characteristics of the<br />

subjects which might conceivably affect the<br />

experimental variables in which the researcher is<br />

interested. If the groups are made equivalent, then<br />

any so-called ‘clouding’ effects should be present<br />

in both groups.<br />

So strong is this simple and elegant true<br />

experimental design, that all the threats to internal<br />

validity identified in Chapter 6 are, according<br />

to Campbell and Stanley (1963), controlled in<br />

the pretest-post-test control group design. The<br />

causal effect of an intervention can be calculated<br />

in three steps:<br />

1 Subtract the pretest score from the post-test<br />

score for the experimental group to yield<br />

score 1.

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

Saved successfully!

Ooh no, something went wrong!