Levels of measurement: Operationalising the DV allows researchers to see if the effect the IV had on
the DV was significant and doing this chooses which level of measurement they want the DV to have.
Nominal- data (qualitative or quantitative ) is put into categories (each participant can only be in one
category) or the frequency is counted. Does not allow for individual comparisons.
Ordinal- data (quantitative) is put into order by asking questions on a scale, giving them a score on a
test or putting them in order like first, second, third. But scales are arbitrary (meaningless when taken
out of the context of the study) and subjective so it doesn't allow for comparisons between
participants.
Interval- data (quantitative) is counted by standardised universally recognised units so comparisons
can be made.
If Unsure, say “at least ordinal”, but it can have multiple answers depending on justifications.
Spearman's Rho: Correlation Ordinal Related
Pearson’s R: Correlation Interval Related
Sign Test: Difference Nominal Related
Chi squared: Difference Nominal Unrelated
Wilcoxon T: Difference Ordinal Related
Mann Whitney U: Difference Ordinal Unrelated
Related T test: Difference Interval Related
Unrelated T test: Difference Interval Unrelated
Related data uses data from the same participants twice, eg a repeated measures design whereas
Unrelated is from independent groups where they're only in one condition
The number of dots on a scattergraph is the number of participants
The correlation coefficient is the calculated value
The sign test: ( Difference Nominal Related )
1) For every participant list their condition A and condition B value
2) Calculate the difference between these two and mark it with 0 + or -
3) S (calculated value) = the amount of + and amount of - and choosing the smallest
4) calculate the critical value (one or two tailed and p=?) but DO NOT include any participants
who got a 0 in the number of participants
5) Conclude
Stats tests conclusions:
To do a stats test we need to know whether it was a directional or non directional hypothesis, the P
value and the number of participants or DF. This then gives the calculated value. This is then put into
a critical value table which shows if it is statistically significant.
The results (ARE/NOT) significant as the calculated value (_) is (HIGHER/LOWER) then the critical
value (_) for a (ONE/TWO) tailed test when n= (_)/ df=(_) and p=(_). This means the experimental
hypothesis has been (SUPPORTED/REJECTED) and the null hypothesis has been
(SUPPORTED/REJECTED) and we can conclude that (_).
Significance is how likely it is that the change in the DV was from the relationship to the IV. Stats tests
simulates repeating the experiment and if the results would be the same each time it is significant,
meaning the aim has been proven. P= is the probability risk (as a decimal) that the results were by
chance. P=0.05 is most likely and is like repeating a study 20 times with a 95% certainty.
A type 1 error is when you wrongly conclude results are significant when they're not and it's caused by
using a P value that is too large (e.g. p=0.1 rather than p=0.01) and not strict enough. A false positive.
To reduce the risk, use a smaller p but higher chance of type 2
A type 2 error is when you wrongly conclude results are not significant when they are and it's caused
by using a p value that is too small and too strict. A false negative.
Using p=0.05 strikes a good balance between making a type 1 or 2 error.
the DV was significant and doing this chooses which level of measurement they want the DV to have.
Nominal- data (qualitative or quantitative ) is put into categories (each participant can only be in one
category) or the frequency is counted. Does not allow for individual comparisons.
Ordinal- data (quantitative) is put into order by asking questions on a scale, giving them a score on a
test or putting them in order like first, second, third. But scales are arbitrary (meaningless when taken
out of the context of the study) and subjective so it doesn't allow for comparisons between
participants.
Interval- data (quantitative) is counted by standardised universally recognised units so comparisons
can be made.
If Unsure, say “at least ordinal”, but it can have multiple answers depending on justifications.
Spearman's Rho: Correlation Ordinal Related
Pearson’s R: Correlation Interval Related
Sign Test: Difference Nominal Related
Chi squared: Difference Nominal Unrelated
Wilcoxon T: Difference Ordinal Related
Mann Whitney U: Difference Ordinal Unrelated
Related T test: Difference Interval Related
Unrelated T test: Difference Interval Unrelated
Related data uses data from the same participants twice, eg a repeated measures design whereas
Unrelated is from independent groups where they're only in one condition
The number of dots on a scattergraph is the number of participants
The correlation coefficient is the calculated value
The sign test: ( Difference Nominal Related )
1) For every participant list their condition A and condition B value
2) Calculate the difference between these two and mark it with 0 + or -
3) S (calculated value) = the amount of + and amount of - and choosing the smallest
4) calculate the critical value (one or two tailed and p=?) but DO NOT include any participants
who got a 0 in the number of participants
5) Conclude
Stats tests conclusions:
To do a stats test we need to know whether it was a directional or non directional hypothesis, the P
value and the number of participants or DF. This then gives the calculated value. This is then put into
a critical value table which shows if it is statistically significant.
The results (ARE/NOT) significant as the calculated value (_) is (HIGHER/LOWER) then the critical
value (_) for a (ONE/TWO) tailed test when n= (_)/ df=(_) and p=(_). This means the experimental
hypothesis has been (SUPPORTED/REJECTED) and the null hypothesis has been
(SUPPORTED/REJECTED) and we can conclude that (_).
Significance is how likely it is that the change in the DV was from the relationship to the IV. Stats tests
simulates repeating the experiment and if the results would be the same each time it is significant,
meaning the aim has been proven. P= is the probability risk (as a decimal) that the results were by
chance. P=0.05 is most likely and is like repeating a study 20 times with a 95% certainty.
A type 1 error is when you wrongly conclude results are significant when they're not and it's caused by
using a P value that is too large (e.g. p=0.1 rather than p=0.01) and not strict enough. A false positive.
To reduce the risk, use a smaller p but higher chance of type 2
A type 2 error is when you wrongly conclude results are not significant when they are and it's caused
by using a p value that is too small and too strict. A false negative.
Using p=0.05 strikes a good balance between making a type 1 or 2 error.
