lOMoARcPSD|22421257
Theme 1-8 - IP 214 year notes
Industrial Psychology
Industrial Psychology (Universiteit Stellenbosch)
214 chapter 4-8
Studocu is not sponsored or endorsed by any college or university
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Theme 4
UNIT 4
What is meant by the requirement that the measure of a measuring instrument should be reliable.
• Reliability= characteristic of a person's behaviour or the performance of an apparatus, which
indicates that one can rely on the person or apparatus [usually in a posive sense] in that the
person or object reacts the same way over me
• If the measures of a human characterisc/trait or of work performance do not remain similar
over several measures of the construct [which is assumed to be stable], it makes no sense to
base any decisions on them; one would not be able to rely on such measures because the
measures do not systemacally point to the [stable] construct to be measured.
• Measuremnt instrument= test, assesment, quesonare
• Construct= e.g. personality, character trait
Formal denions of reliability
• I the measurement doesn’t give us the same scores every me = doesn’t reect the
construct
• Reliability can be formally dened as follows: (you can use any of these denions to dene
reliability)
1) the extent to which a scale tends to render the same measures with
repeated applicaons to the same individual/over repeated measurement
occasions/parallel measures = using the same test on the same person=
similar answers or use the test today and then in a week’s me= similar
outcomes each me or 2 tests prccally the same- the tests should give
similar outcomes
2) the extent to which the variance in the test scores has been caused by
stable, systemacally working inuences= changes in the score reveals
dierences in the construct= one person has more or less of the construct
e.g extroversion
3) the proporon of systemac test variance; in symbols thus ²S/ ²X =
that varaibility in tests= scores are because of systematc inuences that we
expect to see
4) the extent to which the test measurements are devoid of random error
variance [1 – ²e/ ²X.
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Argue with the rst denion as a point of departure (1)
• we do not always get the same score but rather a distribuon of scores in spite of the fact
that that the latent variable remains constant over the measurement period. = no test ever
provides absolute reliability
• This phenomenon is to a certain extent true for any form of measurement.
• It thus follows that no measure is ever 100% / absolutely reliable according to the rst
denion of reliability.
Error distribuon of observed scores
Give you tests repeatedly= you won’t get the same
score every me – your score would be in a bell-
shaped curve – not vercal
Measurement error and error variance
• The queson as to why test measurements are never 100% /absolutely reliable thus changes
into an invesgaon into the causes of the random error score or random error variance=
there are inuences in you as a person and environment that inuence how you respond to
the test
• In order to answer the queson of why some measures are not reliable requires invesgaon
of:
Why people dier in their test performance?
What are the reasons which cause dierences in the scores between dierent tests
or people?
• The basic assumpon of psychological tesng is that any measure contains an element of
error and element of correctness.
• smulus is the test = measuring instrument; your response = we make inference about the
construct we are trying to measure= dierences will be due to your response
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Reasons for individual dierences in test performance
General long-term characteriscs of applicant = you as a person might not have a
certain construct= you not very extroverted
Individual dierences in the characteriscs measured by the test= look at
personality= each individual is dierent and will have their own score or true score
on the construct for that test
Temporary characteriscs of the applicants= not feeling well on test day
Temporary or specic to the test or some part of it= something you don’t understand
in the test or anxiety in tests- specic to the test
Condions of administraon of the test = you don’t like the person administering the
test to you/ venue too hot and noisy
Pure chance= you don’t care so mark everything a
Dierences= seen as error
• Classical psychometric theory assumes that any measure X is the sum of the true measure
(true score) T and a measurement error (error score) E.
X=T+E
• X = observed score = you do a test as get a score- that score is referred to as X
• T=p
roporon true score (reliability of the measure)
• E = proporon error score (unexplained variance)
• The goal of reliability theory is to esmate errors in measurement and to suggest ways of
improving tests so that errors are minimized.
Measurement error
• Assuming that the construct we are measuring will not change over the short-term, the true
score [Ti] is a constant4= whatever your score is= part of that score is actually a true
reecon of the construct we want to measure
• Ei.= if we are looking for a reason for unreliability- it is found in the error
• Keeping in mind that psychological measurement is indirect measurement via a sample of
behaviour, the classical measurement theory acknowledges via Ei the fact that it is virtually
Downloaded by Vianke Herbst ()
Theme 1-8 - IP 214 year notes
Industrial Psychology
Industrial Psychology (Universiteit Stellenbosch)
214 chapter 4-8
Studocu is not sponsored or endorsed by any college or university
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Theme 4
UNIT 4
What is meant by the requirement that the measure of a measuring instrument should be reliable.
• Reliability= characteristic of a person's behaviour or the performance of an apparatus, which
indicates that one can rely on the person or apparatus [usually in a posive sense] in that the
person or object reacts the same way over me
• If the measures of a human characterisc/trait or of work performance do not remain similar
over several measures of the construct [which is assumed to be stable], it makes no sense to
base any decisions on them; one would not be able to rely on such measures because the
measures do not systemacally point to the [stable] construct to be measured.
• Measuremnt instrument= test, assesment, quesonare
• Construct= e.g. personality, character trait
Formal denions of reliability
• I the measurement doesn’t give us the same scores every me = doesn’t reect the
construct
• Reliability can be formally dened as follows: (you can use any of these denions to dene
reliability)
1) the extent to which a scale tends to render the same measures with
repeated applicaons to the same individual/over repeated measurement
occasions/parallel measures = using the same test on the same person=
similar answers or use the test today and then in a week’s me= similar
outcomes each me or 2 tests prccally the same- the tests should give
similar outcomes
2) the extent to which the variance in the test scores has been caused by
stable, systemacally working inuences= changes in the score reveals
dierences in the construct= one person has more or less of the construct
e.g extroversion
3) the proporon of systemac test variance; in symbols thus ²S/ ²X =
that varaibility in tests= scores are because of systematc inuences that we
expect to see
4) the extent to which the test measurements are devoid of random error
variance [1 – ²e/ ²X.
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Argue with the rst denion as a point of departure (1)
• we do not always get the same score but rather a distribuon of scores in spite of the fact
that that the latent variable remains constant over the measurement period. = no test ever
provides absolute reliability
• This phenomenon is to a certain extent true for any form of measurement.
• It thus follows that no measure is ever 100% / absolutely reliable according to the rst
denion of reliability.
Error distribuon of observed scores
Give you tests repeatedly= you won’t get the same
score every me – your score would be in a bell-
shaped curve – not vercal
Measurement error and error variance
• The queson as to why test measurements are never 100% /absolutely reliable thus changes
into an invesgaon into the causes of the random error score or random error variance=
there are inuences in you as a person and environment that inuence how you respond to
the test
• In order to answer the queson of why some measures are not reliable requires invesgaon
of:
Why people dier in their test performance?
What are the reasons which cause dierences in the scores between dierent tests
or people?
• The basic assumpon of psychological tesng is that any measure contains an element of
error and element of correctness.
• smulus is the test = measuring instrument; your response = we make inference about the
construct we are trying to measure= dierences will be due to your response
Downloaded by Vianke Herbst ()
, lOMoARcPSD|22421257
Reasons for individual dierences in test performance
General long-term characteriscs of applicant = you as a person might not have a
certain construct= you not very extroverted
Individual dierences in the characteriscs measured by the test= look at
personality= each individual is dierent and will have their own score or true score
on the construct for that test
Temporary characteriscs of the applicants= not feeling well on test day
Temporary or specic to the test or some part of it= something you don’t understand
in the test or anxiety in tests- specic to the test
Condions of administraon of the test = you don’t like the person administering the
test to you/ venue too hot and noisy
Pure chance= you don’t care so mark everything a
Dierences= seen as error
• Classical psychometric theory assumes that any measure X is the sum of the true measure
(true score) T and a measurement error (error score) E.
X=T+E
• X = observed score = you do a test as get a score- that score is referred to as X
• T=p
roporon true score (reliability of the measure)
• E = proporon error score (unexplained variance)
• The goal of reliability theory is to esmate errors in measurement and to suggest ways of
improving tests so that errors are minimized.
Measurement error
• Assuming that the construct we are measuring will not change over the short-term, the true
score [Ti] is a constant4= whatever your score is= part of that score is actually a true
reecon of the construct we want to measure
• Ei.= if we are looking for a reason for unreliability- it is found in the error
• Keeping in mind that psychological measurement is indirect measurement via a sample of
behaviour, the classical measurement theory acknowledges via Ei the fact that it is virtually
Downloaded by Vianke Herbst ()
