KHA329 Lecture 2: Individual Differences - Psychometrics
1 Jul 2018
School
Department
Course
Professor
Individual Differences week 2: Psychometrics
Measurement, scoring, and norms:
- Measurement: a set of rules for assigning numbers to represent objects, traits,
attributes or behaviours
oMight be intangible
Social anxiety
oAssigning numbers to this
- Psychological tests have rules for assigning a value to an individual’s
performance on that test
oDifferent ways for assigning numbers
- With these numbers we can do things:
oComparison between individual’s scores
- Challenge: psychological properties mostly cannot be measured directly
oStill reliant on the fact that numbers aren’t always right
Try to be as accurate as we can be they will never be absolute
Numbers mean different things in different context
- Raw scores:
oAdding up a performance
oThe numbers we get initially from test responses are raw scores
oRaw scores are the most basic behavioural measurement of an
individual, and are limited in their usefulness
oDon’t have any meaning on their own
oOnly become useful when used for comparisons:
Compared to a reference group of scores
oScores are transformed in carious ways by relating them to
benchmarks, sometimes called norms
oNeed reference points:
Norms based on scores from reference groups
Similar to the person taking the test
- Norms:
oRefer to performance on a test of a particular group called reference
group
oSimple or complex
Eg. Z scores
oNeed appropriate references group
Cultural factors etc.
oNeed to have similar characteristics to the individual being tested
oFactors:
Ethnicity
SES
Gender
Ager
Size and representativeness of normative sample
How many people tested
Are university students really representative of the
normal public
Datedness of norms
Performances change over time due to various reasons
Flynn/Flinn effect: We may look like we are getting
smarter
Understanding of what constructs are are changing
oCreating a norm group:
Random sampling
Finding representative sample
Although some people who take test may not be suitable
Making a test in a certain state, then administering it too
broadly
- What if we have anew test with no information to compare to? Ranking
- Ranking:
oPercentile ranks are important in testing if we want to know where an
individual stands in relation to a reference group
oSimple rank: the position where a case or score lies in order of
magnitude in a group of scores or cases
What if similar scoring
oPercentile rank: the rank or position of a particular case or score
adjusted by the number of scores or cases, expressed as a percentile of
scores falling below that score
Helpful, allow us to adjust simple ranks in terms of where they
fall in a distribution
Calculated percentage of scores
Assuming that people are going ot perform relatively normally
Overall scores will come to normal distribution
Extreme, distorted at the extremes, clumped at either
end rather than center
This can be an issue
The normal distribution and psychological testing:
- If we take a large number of test scores and plot their frequency, they assume
a normal distribution
- Many of the characteristics measured by tests are assumed to be normally
distributed
oIntelligence
- The theoretical normal curve is a very useful instrument in the interpretation
of test scores
- Usually univariate distribution
- Helpful:
oIf a normal curve is perfect, we can figure out exactly how many
people are in a position
- 68% within one SD (above or below)
o2 SDs is extra 13.5%
o3 SDs is further 2.5% (5% of people within 2-3 SD range
- calculate where they are performing relative to everyone else
-
- Z scores:
oStandardized scores are usually symbolized using Z, so are also called
Z scores
oTransform raw scores into standardized units
oShow how distant a raw score is from the mean
oAny raw score can be transformed into a z score by taking the group
mean aay from it then dividing by the standard deviation of the
comparison group
oCan compare different tests
oFancy way of saying SD
oCan create one that is fairly accurate
oIf we know someone’s Z score
All issues around raw scores and means disappears
oA score equal to the mean has SD of 0
Always have an average of 0
oPositive z scores represent raw scores that are above the mean
Negative
- T scores: not all tests use Z
oMean value of 50, SD = 10
oTransfer a ZZ score into a t score: T = 10z + 50
Used in clinical instruments
Psychopathology tests
oIQ: mean = 100, SD = 15
oRemove need for fractions and positive negative signs
Can make it tricky for analysis and interpretation
oT = 10z + 50
oOnly working in whole numbers
oMMPI:
Scales that measure particular characteristics
Eg. Ideas of persecution: ideas about people being out to get
them
Score of 70 = above average
30 = below average
however, this is a test where high or low scores mean
something is going on
Elevated or low scores have clinical meaning
oSocial alienation scale:
SA-1
Scoring high: misunderstood, alienated, others lack in feelings,
lack of support from family, avoid social relationships
A low score on this doesn’t have any clinical meaning
Assessing reliability:
- Reliability: how consistent is the measurement of a construct is
oThe degree to which test scores are free from error
There are alwayf forms of error, so just working at reducing
error
Error with the test itself
Poorly worded items, porr instructions
Not things like; test anxiety, surrounding noise
Document Summary
Social anxiety: assigning numbers to this. Psychological tests have rules for assigning a value to an individual"s performance on that test. With these numbers we can do things: different ways for assigning numbers, comparison between individual"s scores, still reliant on the fact that numbers aren"t always right. Challenge: psychological properties mostly cannot be measured directly. Try to be as accurate as we can be they will never be absolute. Numbers mean different things in different context. Compared to a reference group of scores: scores are transformed in carious ways by relating them to, need reference points: benchmarks, sometimes called norms. Norms based on scores from reference groups. Similar to the person taking the test. Measurement: a set of rules for assigning numbers to represent objects, traits, Norms: group: refer to performance on a test of a particular group called reference, simple or complex. Z scores: need appropriate references group, need to have similar characteristics to the individual being tested, factors:
