Chapter 8: Quantitative Design in Action Research - ar

Download Report

Transcript Chapter 8: Quantitative Design in Action Research - ar

A Short Guide to Action Research
4th Edition
Andrew P. Johnson, Ph.D.
Minnesota State University, Mankato
www.OPDT-Johnson.com
Chapter 8: Quantitative Design in
Action Research
•
Quantitative research is based on the collection and
analysis of numerical data
•
Three quantitative research designs can fit within the
action research paradigm:
1. correlational research
2. causal–comparative research
3. quasi-experimental research
CORRELATIONAL RESEARCH

Seeks to determine whether and to what degree a statistical
relationship exists between two or more variables

Used to describe an existing condition or something that has
happened in the past
Correlation Coefficient
•
Correlation coefficient = the degree or strength of a
particular correlation
•
Positive correlation = when one variable increases, the other
one also increases
•
Negative correlation = when one variable increases, the
other one decreases
•
Correlation coefficient of 1.00 = a perfect one-to-one positive
correlation
•
Correlation coefficient of .0 = absolutely no correlation
between two variables
•
Correlation coefficient of –1.00 = a perfect negative
correlation
Misusing Correlational Research
• Correlation does not indicate causation
• Just because two variables are related, we cannot say that one
causes the other
Negative Correlation
• Increase in one variable causes a decrease in another
Making Predictions
• Correlation coefficient identified by the symbol r
• When r = 0 to .35, the relationship between the two variables is
nonexistent or low
• When r = .35 to .65, there is a slight relationship.
• When r = .65 to .85, there is a strong relationship
CAUSAL-COMPARATIVE RESEARCH

Used to find reason for existing differences between two or more
groups

Used when random assignment of participants for groups cannot
be met

Like correlational research, used to describe an existing situation

compares groups to find a cause for differences in measures or
scores
QUASI-EXPERIMENTAL RESEARCH

Like true experiment; but no random assignment of subjects to
groups

random selection is not possible in most schools and classrooms

Pre-tests and matching used to ensure comparison groups are
relatively similar
Five Quasi-Experimental Designs
•
•
•
•
Exp = experimental group
Cnt = control group
O = observation or measure
T = treatment
Pretest-Posttest Design
Group
Time 
Exp
O
T
O
Cnt
O
—
O
Pretest-Posttest Group Design
Group
Time 
Exp
O
T
O
Cnt
O
—
O
Time Series Design
Group
Time 
Exp
O
Group
Time 
Exp
T1
O
O
O
O
T
O
O
O
O
O
O
O
T2
O
O
O
O
Time Series Group Design
Group Time 
Exp
O
O
O
O
T
O
O
O
O
Cnt
O
O
O
O
—
O
O
O
O
Group Time 
Exp
T1
O
O
O
O
T2
O
O
O
O
Cnt
T1
O
O
O
O
T1
O
O
O
O
Equivalent Time-Sample Design
Group
Exp
Time 
T
O
—
O
T
O
—
O
THE FUNCTION OF STATISTICS
•
Descriptive statistics = statistical analyses used to describe an
existing set of data
•
Measures of central tendency describes a set of data with a single
number
a. mode - score that is attained most frequently
b. median - 50% of the scores are above and 50% are below
c. mean - the arithmetic average
Frequency Distribution = all the scores that were attained and how
many people attained each score
Scores
Number of Students
99
1
97
1
92
2
90
1
85
2
84
4
83
6
80
12
79
5
78
6
75
4
Line graph for frequency distribution
Measures of variability = the spread of scores or how close the
scores cluster around the mean
Range = the difference between the highest and lowest score
Variance = the amount of spread among the test scores
standard deviation = how tightly the scores are clustered around the
mean in a set of data
Scores with a Small Variance
xx
xx
xx
xxx
xx
xx
xx
xx
xx
xx
xx
xx
x
Scores with a Large Variance
x x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Small Standard Deviation: Closely Distributed Scores
Large Standard Deviation: Widely Distributed Scores
INFERENTIAL STATISTICS
• Inferential statistics = statistical analyses used to determine how
likely a given outcome is for an entire population based on a sample
size
• make inferences to larger populations by collecting data on a small
sample size
• Statistical significance = that difference between groups was not
caused by chance or sampling error