Transcript Slide 1 - Pearson

CHAPTER 4
Data:
Measurement and
Analysis
©2005, Pearson Education/Prentice Hall
What to Measure?
• The types of things you can measure in
psychology are endless.
• Deciding on what and how often to measure
can sometimes be a confusing matter.
• One choice you have to make is whether to
measure overt behavior or a covert process.
©2005, Pearson Education/Prentice Hall
Overt Behavior
• Overt behaviors are those that can be viewed or
directly assessed.
• There are 2 types of overt behavior:
– Verbal (use of language, e.g., words)
– Motor (body movement, e.g., running speed).
• Common measures associated with overt
behavior include:
– Performance speed
– Trials
– Reaction time
©2005, Pearson Education/Prentice Hall
Covert Behavior
• Covert behaviors are those that are not
directly observable.
– Some examples include feelings, current
physical states, and attitudes.
– If we can’t directly observe covert behavior
then how do we measure it?
• We can measure covert behaviors directly with
machines (e.g., heart rate monitor) or indirectly
(e.g., self-reports).
©2005, Pearson Education/Prentice Hall
Direct and Indirect Measures of
Covert Behavior
• Direct techniques can include:
– EEG (cortical brain activity)
– GSR (skin conductivity)
– PET and MRI (brain imaging).
• Indirect techniques can include:
– Surveys
– Questionnaires
©2005, Pearson Education/Prentice Hall
Surveys and Questionnaires
• Surveys and questionnaires are important
methods psychologists use to gather information
about covert behavior.
• They generally include some of the following
characteristics:
– Closed-ended items (participants are restricted to a
set of fixed responses, e.g., true/false, multiple choice
questions.)
– Open-ended items (participants are free from
response restriction, e.g., What do you think about
serial killers?).
©2005, Pearson Education/Prentice Hall
Types of Closed-Ended Items
• Closed-ended items are generally superior to
open-ended items because they are much
easier to score (remember this when you do
your Honors Thesis).
• There are many useful ways to present closeended items to participants each with their own
set of pros and cons:
–
–
–
–
Point scales
Likert scales
Segmented rating scales
Numerical rating scales
©2005, Pearson Education/Prentice Hall
Composing a Survey
• Before you start, check the thousands of
published psychological survey and tests
to see if one already exists on your topic.
– If not, then you need to create one. The
following steps will help you:
©2005, Pearson Education/Prentice Hall
Composing a Survey
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Describe, in detail, your measure and refer to this description often.
Decide on who you will measure. Why?
Design items to gather demographic data.
Design numerous items only on the topic you are interested in – do not try to
measure everything.
Rework your items – e.g., make them as simple and straightforward as
possible. Avoid jargon, technical terms, negative wording, and doublebarreled items.
If using closed-ended items, be sure your response sets cover the complete
range of possible answers in equal increments. Why? And be sure you
response sets match the question. Why?
Decide on the statistics you are likely to use.
Pretest the survey on a group of people who will not be in your actual study.
Tell them you want constructive criticism.
Revise the test and make it look professional.
Now the fun part begins – giving the survey and deciding if your survey is
reliable and valid. This comes in later chapters.
©2005, Pearson Education/Prentice Hall
Other Types of Data: Remnant
• Remnant data refers to the use of existing
remains, products, or evidence of behavior
to infer or explain past events.
• Some examples include:
– Physical Traces
• E.g., Graffiti, garbage.
– Archival Data
• E.g., birth weights and rates, weather reports and
crime.
©2005, Pearson Education/Prentice Hall
How to Identify Pseudoscience?
• The “Trappings” of Science
– Pseudoscience tries to be like real science.
– Pseudoscience makes predictions about
phenomenon, but rarely tests them.
• Data is Often Based on Testimonials
– Data like this can be easily manipulated.
• Evasion of Disproof
– Explanations given by the pseudoscience to account
for data that disprove the pseudoscience are
difficult/impossible to test (e.g., the phenomenon can’t
be measure by conventional means)
©2005, Pearson Education/Prentice Hall
Naturally Occurring Behavior
•
•
•
Sometimes psychologists want to study
ongoing or naturally occurring behavior.
This is unique because researchers not only
have to be concerned about what they are
observing, but also when and where the
behavioral observations will be made.
To deal with this, psychologists often employ:
– Time sampling technique
• Taking samples of behavior only at certain times.
– Event or Situation sampling technique
• Taking samples only in a predetermined situation.
©2005, Pearson Education/Prentice Hall
Types of Data
• The are two main categories of data:
1. Quantitative
• Is numerical or can easily be converted to
numerical form.
2. Qualitative
• Usually narrative in nature and difficult if not
impossible to convert to numerical.
©2005, Pearson Education/Prentice Hall
Scales of Measurement
• Once you have determined the type of data you
will generate, you now need to coordinate this
with statistics procedures.
• The first step in doing this is to determine your
data’s “scale of measurement”.
– There are 4 Different Scales of Measurement
1.
2.
3.
4.
Nominal
Ordinal
Interval
Ratio
Let’s consider each one.
©2005, Pearson Education/Prentice Hall
Nominal Scale
• Most basic level of measurement.
• Numbers represent simple qualitative
differences in your variables.
– E.g., 1 = group_1; 2 = group_2, etc.
• Numbers are not intended for numerical
calculations, but to classify data.
• General rule: Similar objects or events are
assigned similar numbers, and different objects
get different numbers.
• Statistical procedures:
– Frequency counts; Chi-Square.
©2005, Pearson Education/Prentice Hall
Ordinal Scales
•
Numbers in ordinal scales describe the object or event as they did
in nominal scales, but they also assign magnitude in the form of
rank or order.
•
Ordinal scales indicate an individual’s or object’s value based on its
relationship to others in the group. Thus, the numbers have
meaning only within the group.
•
Ordinal scales provide no information about how closely two
individuals or objects are related. Thus, you cannot add, subtract,
multiple, or divide numbers in an ordinal scale.
–
You can transform ordinal numbers as long as the original information
about rank is preserved.
•
E.g., Olympic medals, rank of professors.
•
Appropriate statistics: percentiles, correlation, Mann-Whitney U.
©2005, Pearson Education/Prentice Hall
Interval Scale
•
•
•
•
•
•
Numbers are assigned with the assumption that each
number represents a point that is an equal distance
from the points adjacent to it.
An interval scale is thus very much like the number
system you are familiar with.
Interval scale also have all the properties noted in
nominal and ordinal scales.
Interval scales do not have true zeros, thus you cannot
make ratio comparisons.
E.g., Fahrenheit temperature scale, many Likert
scales.
Statistical procedures: mean, median, mode, standard
deviation, correlation, t-test, ANOVA.
©2005, Pearson Education/Prentice Hall
Ratio Scale
•
•
•
•
All the properties of the preceding scales, plus
a true zero.
E.g., Kelvin temperature scale, bathroom
scale.
Since true zeros are present ratio comparisons
can be made (90 kg is exactly 3 times more
than 30 kg).
Appropriate statistics: mean, median, mode,
standard deviation, correlation, ratios, t-tests,
ANOVA.
©2005, Pearson Education/Prentice Hall
Descriptive Statistics
•
•
Descriptive stats provide information about the
central tendencies of a group of data.
Importance terms include:
–
–
–
–
Mean: the arithmetic average in a data set
Median: the middlemost score in a data set
Mode: the most frequent score in a data set
Variance: the degree to which scores in a data set
deviate from the mean. There are various ways to
measure variance or variability.
• Range: measure of variability in a data set
• Standard Deviation: most commonly used measure of
variability.
©2005, Pearson Education/Prentice Hall
Inferential Statistics
• Inferential statistics are mathematical /
statistical procedures for determining the
probability that the relationships or differences
we observe in our data actually occur in the
population.
• Inferential statistics also tell us whether the
differences we see in our data occurred by
chance or not.
• There are 2 types of inferential statistics:
– Parametric statistics
– Nonparametric statistics
©2005, Pearson Education/Prentice Hall
Parametric and Nonparametric
Statistics
• Parametric Statistics test hypotheses that are
based on data that allow us to estimate
parameters (e.g., means and standard
deviations).
– In other words, parametric statistics are used with
interval or ratio data.
– E.g. Pearson r, multiple regression, t-test, ANOVA
• Nonparametric Statistics test hypotheses that
do not involve parameters (e.g., when the data
are nominal or ordinal, or not normally
distributed).
– E.g., Spearman rank correlation, Chi-square.
©2005, Pearson Education/Prentice Hall
Statistical Significance
•
What do we mean when we say something is
statistically significant?
– We are simply saying that there is only a small probability that
what we found was due solely to chance.
– That small “chance” probability goes by various names:
• Type I Error
• Alpha level
• And is often symbolized by an italicized p “p”
•
The alpha level is predetermine before the study begins
and in psychology the level is usual set to 5% or 0.05.
– This means that if the results we gather cannot be obtained
by chance more that 5 times in 100 random trials we would
say that our results are statistically significant. It is unlikely
they occurred solely by change. So can we ever be wrong?
©2005, Pearson Education/Prentice Hall