PSYC 3020 EXAM QUESTIONS WITH
ANSWERS 100% CORRECT
Name the scales of measurement and define them - ANSWER Nominal: names,
categorical and dichotomous data
Ordinal: Rank order with no meaningful measurements between things
Interval: Distances between things are meaningful but zero is arbitrary
Ratio: Ratio between things is meaningful and true zero means the absence of the thing
Categorical data can always be measured on a ___ scale - ANSWER nominal
Which scales of measurement must use a parametric test? - ANSWER Interval and Ratio
Which scales of measurement must use a non-parametric test? - ANSWER Nominal and
Ordinal
In a normal distribution how many scores are +/- 1 standard deviation around the mean ?
- ANSWER 68%
In a normal distribution hoe many scores are +/- 2 standard deviations around the
mean? - ANSWER 95%
Why would we normalise a distribution? - ANSWER Normalising a distribution makes
scales comparable and we are able to perform more powerful tests on the data
(parametric)
What are the options to normalise a distribution? - ANSWER 1. perform a non-linear
transformation. This doesn't change the rank-order, but stretches some bits of the scale
more than others
2. Redesign the measure if possible (i.e. change the wording of the questionnaire)
3. Leave as is and use a non-parametric test
How do we find a z score? - ANSWER _
x-x
----
Sx
x take the mean divided by the standard deviation
,How do we find percentile ranks ? - ANSWER First transform into z scores. Then look up
the table of a normal distribution. if z>0 use larger portion. If z<0 use smaller portion.
Is converting z scores into percentile ranks an example of linear or non-linear
transformation? Why? - ANSWER Non-linear transformation. This is because it changes
the shape of the data.
What do correlation coefficients tell us. - ANSWER Correlation coefficients are
measures of effect size. They tell us the strength of a relationship but not whether it is
statistically significant.
How does scatter effect the correlation coefficient? - ANSWER The more scatter there
is, the lower the correlation coefficient will be.
What does the magnitude of the correlation coefficient tell us about the slope? -
ANSWER Nothing. The magnitude of a correlation coefficient does not reflect the slope
of linear relationships between two variables, scatter does.
Why is variation important to correlations? - ANSWER Because we are measuring the
variation. If the spread is truncated, then the correlation coefficient magnitude nay be
reduced.
Give an example of why variation is important - ANSWER To avoid ceiling effects, i.e. if
you include novice and expert electricians but make the task very easy such as
changing a lightbulb you won't be able to find a significant difference between the
novices and experts.
What are margin of errors in significance testing for correlation? - ANSWER Also known
as confidence intervals, mean around the answer which deviates from being accurate.
This is determined by sample size.
Types of behavioural research designs that can be used in program evaluations. -
ANSWER Experiments (testing program outcomes using a randomised control trials)
Quasi- or natural experiment (measuring program outcomes but without control group is
not randomly assigned)
Quantitative Survey
Qualitative research (focus groups, interviews, etc)
What is psychometrics? - ANSWER The science of psychological measurement
What are the different types of psychological tests? - ANSWER Ability (right and wrong).
, Includes achievement tests (testing something you learnt) and aptitude tests (tests
general ability)
Personality based tests (no right or wrong answer)
Explain norms when standardising scores - ANSWER When standardising scores we do
it relative to a normative sample. This may include age norming (choosing samples of
specific ages which relate to your study). The larger the sample is the more stable it will
be and won't be as affected by sampling error.
The Law of Large Numbers - ANSWER As a sample size grows, it's mean gets closer to
the average population
What is stratified cluster sampling? - ANSWER Deliberately recruiting certain individuals
and groups to get the desired ratio in the sample to achieve representativeness
What else can we do other than stratified cluster sampling to achieve the desired ratio in
a sample? - ANSWER We can weight individuals or groups accordingly to give them
equal representation. For example, a sample has a ratio of boys outnumbering girls by
1.6, girls can be weighted equally to them to make their scores count for the equivalent
amount.
Another strategy we can use is randomisation
Name the standardised scores - ANSWER z scores
percentile ranks
t scores
IQ scores
How do we convert a z score to a t score ? - ANSWER multiply by 10 (SD) and add 50
(Mean)
ANSWERS 100% CORRECT
Name the scales of measurement and define them - ANSWER Nominal: names,
categorical and dichotomous data
Ordinal: Rank order with no meaningful measurements between things
Interval: Distances between things are meaningful but zero is arbitrary
Ratio: Ratio between things is meaningful and true zero means the absence of the thing
Categorical data can always be measured on a ___ scale - ANSWER nominal
Which scales of measurement must use a parametric test? - ANSWER Interval and Ratio
Which scales of measurement must use a non-parametric test? - ANSWER Nominal and
Ordinal
In a normal distribution how many scores are +/- 1 standard deviation around the mean ?
- ANSWER 68%
In a normal distribution hoe many scores are +/- 2 standard deviations around the
mean? - ANSWER 95%
Why would we normalise a distribution? - ANSWER Normalising a distribution makes
scales comparable and we are able to perform more powerful tests on the data
(parametric)
What are the options to normalise a distribution? - ANSWER 1. perform a non-linear
transformation. This doesn't change the rank-order, but stretches some bits of the scale
more than others
2. Redesign the measure if possible (i.e. change the wording of the questionnaire)
3. Leave as is and use a non-parametric test
How do we find a z score? - ANSWER _
x-x
----
Sx
x take the mean divided by the standard deviation
,How do we find percentile ranks ? - ANSWER First transform into z scores. Then look up
the table of a normal distribution. if z>0 use larger portion. If z<0 use smaller portion.
Is converting z scores into percentile ranks an example of linear or non-linear
transformation? Why? - ANSWER Non-linear transformation. This is because it changes
the shape of the data.
What do correlation coefficients tell us. - ANSWER Correlation coefficients are
measures of effect size. They tell us the strength of a relationship but not whether it is
statistically significant.
How does scatter effect the correlation coefficient? - ANSWER The more scatter there
is, the lower the correlation coefficient will be.
What does the magnitude of the correlation coefficient tell us about the slope? -
ANSWER Nothing. The magnitude of a correlation coefficient does not reflect the slope
of linear relationships between two variables, scatter does.
Why is variation important to correlations? - ANSWER Because we are measuring the
variation. If the spread is truncated, then the correlation coefficient magnitude nay be
reduced.
Give an example of why variation is important - ANSWER To avoid ceiling effects, i.e. if
you include novice and expert electricians but make the task very easy such as
changing a lightbulb you won't be able to find a significant difference between the
novices and experts.
What are margin of errors in significance testing for correlation? - ANSWER Also known
as confidence intervals, mean around the answer which deviates from being accurate.
This is determined by sample size.
Types of behavioural research designs that can be used in program evaluations. -
ANSWER Experiments (testing program outcomes using a randomised control trials)
Quasi- or natural experiment (measuring program outcomes but without control group is
not randomly assigned)
Quantitative Survey
Qualitative research (focus groups, interviews, etc)
What is psychometrics? - ANSWER The science of psychological measurement
What are the different types of psychological tests? - ANSWER Ability (right and wrong).
, Includes achievement tests (testing something you learnt) and aptitude tests (tests
general ability)
Personality based tests (no right or wrong answer)
Explain norms when standardising scores - ANSWER When standardising scores we do
it relative to a normative sample. This may include age norming (choosing samples of
specific ages which relate to your study). The larger the sample is the more stable it will
be and won't be as affected by sampling error.
The Law of Large Numbers - ANSWER As a sample size grows, it's mean gets closer to
the average population
What is stratified cluster sampling? - ANSWER Deliberately recruiting certain individuals
and groups to get the desired ratio in the sample to achieve representativeness
What else can we do other than stratified cluster sampling to achieve the desired ratio in
a sample? - ANSWER We can weight individuals or groups accordingly to give them
equal representation. For example, a sample has a ratio of boys outnumbering girls by
1.6, girls can be weighted equally to them to make their scores count for the equivalent
amount.
Another strategy we can use is randomisation
Name the standardised scores - ANSWER z scores
percentile ranks
t scores
IQ scores
How do we convert a z score to a t score ? - ANSWER multiply by 10 (SD) and add 50
(Mean)
