Inferring associations Flashcards
(25 cards)
what does an association mean?
An association between two measured concepts looks at whether there is a tendency for the scores on two variables to change in a common/linked manner.
when would you use a Pearson correlation?
looking for an association
continuous variables
normally distributed
parametric test
if you are looking for an association between 2 continuous variables that are normally distributed, you would use a…
Pearson correlation
how does a Pearson correlation work?
Pearson Correlation Coefficient (r) estimates an association as a ‘linear relationship’
Scores of r range from -1 (perfect negative correlation) to 0 (no correlation) to +1 (perfect positive correlation)
null hypothesis for a 2-tailed Pearson Correlation test:
there is no correlation between the variables (that r=0) in the population.
SPSS output and report for a Pearson correlation
Is the correlation coefficient positive or negative?
Is it likely that the correlation is untrue in the population? (p<0.05)
example write-up of the results from a Pearson correlation
There was a positive statistically significant correlation between a person’s birthweight and their attainment in maths at age 10 years (r(690)=0.09, p<0.05)
What information is required to carry out a Pearson correlation?
Pearson correlation coefficient, significance and number of participants
when would you use a Spearman correlation?
looking for an association
continuous variables that are not normally distributed
OR
2 ordinal variables
OR
1 ordinal 1 continuous variable
if you are looking for an association between continuous variables that are not normally distributed, you would use a…
Spearman correlation
if you are looking for an association between 2 ordinal variables, you would use a
Spearman correlation
if you are looking for an association between a continuous and ordinal variable, you would use a…
Spearman correlation
how does a Spearman correlation work?
Turns the scores of the continuous and ordinal variables into ranks
Then analyses this rank data instead of the original variables (discarding some information makes it less desirable to use)
Spearman’s correlation (ρ) estimates the association as a ‘linear relationship’
Scores of ρ range from -1 (perfect negative correlation) to 0 (no correlation) to +1 (perfect positive correlation)
null hypothesis for a 2-tailed Spearman’s correlation test
‘there is no correlation between the variables (that ρ=0) in that population’
SPSS output and report for Spearman’s correlation
Is the ‘correlation coefficient’ positive or negative?
Is it likely that the correlation is untrue in the population? (p<0.05)
Example write-up of the results from a Spearman correlation:
‘there was a positive statistically significant correlation between a person’s birthweight and their attainment in maths at age 10 years (ρ(690)=0.11, p<0.01)’
What information is required to carry out a Spearman correlation?
Spearman correlation coefficient, significance and number of participants
when would you use a chi-square test?
when looking for an association between 2 nominal variables
OR
1 ordinal and 1 nominal variable
if you are looking for an association between 2 nominal variables, you would use a …
Chi-square test
if you are looking for an association between an ordinal and nominal variable, you would use a…
chi-square test
how does a chi-square test work?
Counts how many participants have each combination of the combined categories within your non-continuous variables
Considers whether these are ‘as expected’ meaning there is a balanced proportion across categories
If there are more participants ‘observed’ in a category than we would ‘expect’ then this is evidence of an association
null hypothesis for a 2-tailed chi-square test:
‘within the underlying population, there will be an equal number of participants counted within the combined categories as there are ‘expected’
SPSS output and report for Chi-Square test
Where are there greater and fewer ‘counts’ than the ‘expected counts’
Are any ‘adjusted residuals’ outside of the -1.96 to +1.96 range?
Is it likely that the observed counts will equal the expected counts in the population? (p<0.05)
Example write-up of the results from a chi-square test:
“there was a statistically significant association between a mother’s school leaving age and her 10 year old child’s likelihood of receiving Free School Meals in 1980 (Χ²(1, n=706) = 22.22, p<0.001). The receipt of Free School Meals was more likely for children aged 10 years whose mothers had left school aged under 16 years”
