Repeated Measures Design and Correlations Flashcards
What is a repeated measures design?
Participants are tested in 2 conditions (or more)
What are the advantages of repeated measures designs?
> Fewer participants are needed to run an experiment as compared to a between-groups design
This can save material, money and effort
Less error in your experiment than there would be if you had 2 separate groups
Disadvantage of repeated measures designs?
Practice/ Order effects
what factors can practice effects occur due to?
A number of different factors e.g. learning, fatigue, adaptation, habituation, sensitisation, contrast etc.
How do we combat practice/order effects?
COUNTERBALANCING
What is counterbalancing?
This means assigning participants to different conditions in different orders.
What is complete counterbalancing?
Takes into account every possible combination e.g. 123, 231, 312.
when does complete counterbalancing become more difficult?
when you have lots of different conditions- lots of different combinations.
What is partial counterbalancing?
randomly aelecting a sub set of possible conditions orders and assigning participants randomly to those.
One type of ______ design is called a correlational design.
Repeated measures
How many variables will a correlational design have?
2
What scales can variables be on in a correlational design?
either ordinal, interval or ratio.
What study design is this?
Increasing self harm is associated with lower self esteem.
Correlational design
A correlation is a description about how…
2 variable are associated with one another
name the 2 types of correlation.
Positive and Negative.
name this correlation type:
As one variable increases so does the other.
Positive :)
name this correlation type:
As one variable increases, the other decreases.
negative.
What is the descriptive statistic for a correlation?
A scatterplot :)
If you have an IV and a DV, what axis should the IV be on in a scatter plot?
The x-axis.
what can we draw to help us see a potentially significant trend on the scatter plot?
A line of best fit
What can the line of best fit show us?
A potentially sig relationship between 2 variables.
Also helps identifying outliers.
If correlation = 0 what does this mean?
There’s no correlation.
If correlation = -1 what does this mean?
Strong negative correlation
If correlation = +1 what does this mean?
Strong positive correlation
the stronger the correlation, the ____ the dots to the line of best fit.
closer
what does r stand for?
correlation :)
what is the strongest correlation you can get?
r= 1
OR
r = -1
What does r = 0 mean?
No linear relationship
If there is no linear relationship what should we consider?
There might still be a non-linear relationship e.g. increase then decrease.
What is an outlier?
A really unusual score
There are 2 options re- what we do with our outliers?
delete it
OR
keep it and report the fact there is outliers.
It’s probably best to do what with outliers?
delete them
You can use a scatterplot to identify outliers, but what else could you use?
SPSS
How do we identify outliers on SPSS?
If they are greater than or less than 3.29.
Any significant correlation has a number of interpretations- what are these?
- We made a type I error
- Changes in A caused change in variable B
- Changes in B caused change in variable A
- Both changes in variables A and B influence each other
- the 3rd variable problem- and omitted variable C accounts for the link between A and B!
Remember correlation is not ________.
causation
don’t say A causes B- what should we say instead?
There’s a sig. relationship between A and B
What visually represents our correlation?
scatterplot
what is the coefficient determination?
literally jus the correlation multiplied by itself e.g. r squared.
Caleculate the effect size:
r= .5
r squared = .25
What does an effect size of .25 tell us?
25% of the variation in A can be accounted for by variation in B (or vice- versa).
The left over percentage from the effect size .25 means what?
means that 75% of variation is due to random error and other things we haven’t measured.
If effect size = 1 we could perfectly predict the ……
we couldn’t predict it at all of effect size was ___.
second variable
0
what is a small effect size?
r2 = .01 to .08
what is a medium effect size?
r2 = .09 to .24
what is a large effect?
r2= .25 and above.
What do parametric tests assume about your data?
> interval or ratio data
normally distributed- not skewed.
at least 10 participants on each level of you dependent variable.
What is the parametric test for correlations?
Pearson Correlation Coefficient
aka. Pearsons
What is the non-parametric test for correlations?
Spearman Rank Order Correlation
aka. Spearman’s Rho
What is a one tailed hypothesis?
It states the direction of the correlation i.e. “there will be a significant positive correlation between…”
What is a 2 tailed hypothesis?
It doesn’t give a direction- it just says “there is a significant correlation between the 2 variables”
Pearson’s correlation revealed a significant, ______ correlation between self esteem and depression, r = -.88, n=54, p=.020.
What is the effect size?
negative
effect size = r2= .77
The alternative hypothesis takes 2 different forms, what are they?
1 tailed and 2 tailed.
in a 2 tailed hypothesis, what happens to the 5%?
It’s split 2.5% either side.
In a 1 tailed hypothesis, what happens to the 5%?
The 5% is all to the one side.
What is good about a 1 tailed?
Greater chance of accepting something as significant IF it’s in the direction you accepted.
what data does Spearman’s Rho require?
at least ordinal
Rs stands for….
Spearman’s Rho
R s = .94, n=10, p=
NOTHING- no effect size reported in Sperman’s Rho.
What are the 2 ways to report p values?
p = .xxx
OR
P < .001
How many dp is p reported to?
3!
Correlation can be used to establish ____ reliability.
test
What is this called?:
Do same test to same sample of participants at 2 different times e.g. 1 week apart. Then correlate their scores on two sets of responses, if it’s a reliable measure then the 2 sets of scores should be highly correlated (>.70). Reported using standard format as a normal correlation (no need for effect size).
Test reliability
What is split half reliability?
Add up scores from half test items.
Add up other half.
Both scores are then correlated, if both halves are measuring the same construct they should be highly correlated (> 0.70). rsb = .78 is how it’s reported.
What is internal reliability also known as?
Cronbach’s alpha
____ _____ represents the average degree to which each item is correlated with all other items.
Cronbach’s Alpha
What should Cronbach’s alpha be above to indicate a reliable scale?
.70
What does Cronbach’s alpha test?
Basically of all items on the scale are measuring the same thing.