Selecting a test: lesson 2

advertisement
Selecting a test: lesson 2
Choice 3 Levels of data
1st Choice Difference or Correlation
• The first decision you have to make is
what the hypothesis is testing. Is it talking
about a difference between two conditions
or is it asking about the relationship
between two variables.
• It’s all in the language used.
Revisiting the past
• Recall the following experiments that you
have studied, how are the participants
measured.
• Primacy/Recency
• Milgram’s obedience study
• Asch’s conformity study
Different types of data
• For the purposes of statistical analysis,
there are different types, or levels, of
data. On this basis, different inferential
tests work on different levels of data.
Data can be collected on 3 levels.
• Nominal
• Ordinal
• Interval
• If a researcher has categorised a
participant in some way then it is nominal
data. However, if each participant has
some kind of score then it is interval or
ordinal.
Have a look at these 2 data tables.
The Distribution of respondents
answers to the statement ‘ More
criminals should be treated through
rehabilitation rather than
imprisonment’.
Strongly
agree
Agree
12
Disagree
33
Strongly
Disagree
15
27
Participants’ raw scores in a test
of memory
Participant
1
2
3
4
5
6
Score
11
9
11
15
7
11
Categorising Data
• In small groups, devise a categorising
system to divide the class up. Think about
what will define each group and how this
will be measured.
• Feedback
Nominal Data
• Nominal data are in categories. If a person
was categorised as tall or short, this would
give us nominal data on them. Other
examples of nominal data include
classifying people as introvert or extrovert;
identifying people as high, moderate or
low achievers; observing whether people
did or did not stop at traffic lights.
• Essay grades can be nominal data. By
deciding whether a student has passed or
failed an essay, teachers are putting their
performance into a category. This also
illustrates why nominal data is the lowest
level of data-because it gives us limited
information on the variable being
measured.
• In the case of an essay, the variable is a
student’s understanding of a topic. What
indication does pass or fail give?
What are the
problems
with nominal data
Which one of these experiments, if
any, is nominal level data…….why?
• Primacy/Recency
• Milgram’s obedience study
• Asch’s conformity study
Ordinal Level
• This is a
measurement scale in
which the scores are
ranked.
• For example the order
of the finishers of a
race – first, second,
third and so on.
• Importantly no information is provided
about the size of the difference between
any two scores.
• For example if we know the order of the
finishers of a race this does not tell us
anything about the time difference
between the racers.
• Produce some ordinal data to describe the
class.
Interval Ratio Level
• This is a measurement scale with equal units of
•
•
intervals throughout the scale.
For example, the times in seconds the racers
finished the race.
If the measurement scale does not have an
absolute zero we call this interval data. When
the lowest point on a scale is zero and the units
of intervals are equal we call this interval ratio
data.
• If the measurement scale does not have
an absolute zero we call this interval data.
• When the lowest point on a scale is zero
and the units of intervals are equal we call
this interval ratio data.
• The distinction does not matter when we
are choosing a test.
• Produce some examples of interval level
data.
Examples
• Researchers want to find the difference between
•
•
male and female estimates of stopping
distances.
On a questionnaire about social behaviour there
is a question asking: On a scale of 1-10 how
aggressive is your child? 1 is not at all 5 average
and 10 very aggressive
Participants have to choose ‘snog’, ‘marry’ or
‘avoid’ when shown a set of photographs of
possible partners
Answers
• Interval
• Ordinal
• Nominal
The last choice:
Related or Unrelated
• Related samples (e.g. repeated measures
design or matched pairs design).
• Unrelated samples (e.g. independent
measures design).
• This is very simple knowledge from AS.
Try The Card Sort
• Test your knowledge of the experimental
designs.
• Match the strengths and weaknesses to
the designs.
Download