Class Notes
(808,131)

Canada
(493,086)

York University
(33,532)

Psychology
(4,070)

PSYC 2230
(205)

N/ A
(9)

Lecture 6

# RESEARCH METHODS LECTURE 6.pdf

Unlock Document

York University

Psychology

PSYC 2230

N/ A

Summer

Description

RESEARCH METHODS
LECTURE 6: JULY 12TH, 2012
TOPIC: MEASUREMENTAND STATISTICALANALYSIS/ EXAM REVIEW
SCALES OF MEASUREMENT
➔ NOMINAL → differences between nominal categories are QUALITATIVE instead of
QUANTITATIVE – no numerical difference
➔ Example → Males Versus Females Introvert versus Extrovert Sport Car versus SUV
➔ Individuals can only be in ONE category and not MULTIPLE categories
➔ Some people feel this scale is too primitive
➔ Summarizing any data and analyzing is limited
➔ Can only make qualitative differences (whether two people are the same or different)
➔ Report frequency in the form of percentages or proportion frequencies
➔ ORDINAL→ involve rank ordering st nd rd
➔ Example: results of a horse race (1 , 2 , 3 etc ) , class rank
➔ Can determine whether two individuals are different as well as the direction of difference
(example: someone who places 4 versus someone who places 2 ) nd
➔ No indication of the magnitude of difference
➔ INTERVAL → rank ordering and the distance between categories is assumed to be identical
➔ Example: temperature between 40 and 50 degrees is the same as 70 and 80 degrees
➔ -2 -1 0 1 2 horrible to awesome
➔ With Likert scales it is assumed that the distance between individual points on the scale is
identical
➔ Likert scales → how would you rate the quality of Eugene winter weather?
➔ It allows us to calculate the absolute distance between items as well as the average across a
number of items but it does not allow us to make conclusions about the ratio of one value to
another. Cannot say 30 degrees Fahrenheit is 3 times as warm as 10 degrees Fahrenheit.
➔ RATIO → rank ordering with a constant difference between adjacent categories and there is an
absolute zero point
➔ Examples: height, distance, reaction time
➔ The existence of an absolute zero point allows calculations of ratios between different values. 6
feet is 2x as long as 3 feet
➔ Each of these measures has a zero point that corresponds to the absence of the measured item
➔ Categorical or Discrete → nominal and ordinal variables
➔ Continuous → interval and ratio variables
➔ Nominal → least amount of info
➔ Ordinal → adds some info (quantity and rank)
➔ Interval and ratio give a lot of info overall
➔ Statistical tests available → tests for nominal and ordinal data (nonparametric—such as chi
square) are less powerful than those available for interval and ratio data (parametric—such as T-
test or ANOVA)
➔ Psychometric Properties → referring to the reliability and validity of a psych test (it is the
quality of the test)
➔ This is established typically over time. It is never usually established by one single study.
➔ Agood test should be reliable → consistent, predictable, and valid → measures what it intends to measure
RELIABILITY
➔ Consistency or stability of a measure or of behaviour
➔ Produces similar results when repeated measurements are made under identical conditions
➔ Observed score → true score + error
➔ Sources of error → observer error (human error in process), environmental error (time of day,
noise level, weather conditions, lighting), participant changes (boredom, hunger, mood) →
subtle changes
➔ Assumed that the error component is the factor responsible for the changes from one
measurement to the next
➔ Error component relatively small, your scores will be relatively consistent (reliable)
➔ If the error component relatively large, you will find huge differences from one measurement to
the next (not reliable)
➔ The error is what changes (ex. Person is currently depressed, happy → participant changes)
➔ Unreliable measures hinder the research process because the results you would obtain would be
unstable and you may not be able to replicate them, would not be able to make accurate
conclusions
➔ We would obtain different each time we measure the same behaviour → leading to different
conclusions each time
➔ Measured using a correlation coefficient
TYPES OF RELIABILITY
➔ (1) Test -Retest → give the same test to the same people at multiple time points and keep all the
conditions the same or else you would introduce error. Correlating time 1 scores with the time 2
scores. The more time that elapses the lower the test-retest reliability becomes
➔ Problem? They might remember or recall the items and the second score might go way up
simply because they remember how they responded the first time. (Correlation may become
artificially high)
➔ (2) Alternate/Parallel Form → Give two different but equivalent forms to the same people at
two different points in time. Compare the scores. Example: split the sample into two parts: 1
and 2
➔ Version 1 now version 2 six months later (they may give even numbers first then odd numbers
as a strategy so you cannot recall).
➔ Problem? They may learn something the first time around, it might be difficult a parallel or
alternate form (some items the second time around may be a greater or lower difficulty level
compared to the first time around).
➔ Cost: constructing an alternate form parallel to the original one may be hard
➔ (3) Internal Consistency → Takes advantage of a test itself. It assesses reliability from the
statistical interrelationships amongst the actual items (responses) of the test to other items on
the test (scores on separate parts of the test).
➔ To measure it (types of internal consistency 1): split-half method → arbitrarily splitting a test
into two equal components and then correlating part 1 with part 2. Look at part 1 with part 2. or
maybe take odd items and correlate them with the even items. Correlation between 2 halves
should approach one if it is reliable.
➔ (Types of internal consistency 2) CronbachAlpha → calculating the correlation of each item
with every other item in that scale. CronbachAlpha is the average of all the correlation coefficients. If very low, then it means the items have very little in common
➔ Y

More
Less
Related notes for PSYC 2230