Statistical Analysis for Education and Psychology Researchers

Multiple Samples (groups) More than 1 independent variable (factors)

Research Q:

Differences

between

groups (class

variable is

discrete

forming

groups)

Cochran’s

test for>2

proportions

(6.7) Q

— — Two-way

ANOVA

(unrelated)(

(8.9)

Two-way

ANOVA

split-

plot’)

(8.10)

Association

(class variable

is discrete

forming

groups)

r×k χ^2 test

(6.6)

— — — — —

* Can be independent or related; (6.1) indicates Chapter and section in text; † Independent

and related

Figure 5.1: Decision chart for

choosing a statistical test

ables, continuous or discrete variables are also stated under the design column. Discrete
independent variables are sometimes called class variables because they classify subjects
into relevant groups for analysis.
You should then follow the row along and stop when the row intersects with an
appropriate column heading relating to type of data and inference, for example,
binomial/nominal; rank or normal data distributions. For some designs you need to
choose between an independent or related measure for the response variable and also
make a choice between discrete or continuous independent variables. At the beginning of
each of the following chapters there is a summary of statistical tests included in that
chapter.

Example 5.1: Choosing an Appropriate Statistical Test

In part of a study designed to identify whether there were any differences between
mathematics and verbal self-concept among Norwegian school children, Skaalvik and
Rankin (1994), sampled 165 sixth grade boys and 191 sixth grade girls. An additional
question addressed in the study was whether any differences in self-concept were gender
related. Two of the scales developed and used in the study were the Mathematics Self-
Concept (MSC) scale; self-perceived ability to learn mathematics, for example one item
is ‘I have high mathematics aptitudes’, and the Verbal Self-Concept (VSC) scale; verbal
self-concept about mathematics, an example of one item is, ‘I have no problems learning
mathematics.’ Means and standard deviations were presented for both scales.
In choosing a statistical test the first consideration is,

What is the general research design, how many groups (measures) are

involved?

Statistical analysis for education and psychology researchers 128