A2 RESEARCH METHODS - STATS TESTS, PROBABILITY AND SIGNIFICANCE Flashcards
What stats test would be used for an IG design and nominal data?
Chi-squared (OV>/=CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for an RM/MP design and nominal data?
Sign test (OV less than or equal to CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for a correlation and nominal data?
Chi-squared (OV>/=CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for an IG design and ordinal data?
Mann-Whitney U (OV less than or equal to CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for an RM/MP design and ordinal data?
Wilcoxon (OV less than or equal to CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for a correlation and ordinal data?
Spearmans rho (OV>/=CV)
THIS IS A NON-PARAMETRIC TEST
What stats test would be used for an IG design and interval/ratio data?
Unrelated t-test (OV>/=CV)
THIS IS A PARAMETRIC TEST
What stats test would be used for an RM/MP design and interval/ratio data?
Related t-test (OV>/=CV)
THIS IS A PARAMETRIC TEST
What stats test would be used for a correlation and interval/ratio data?
Pearsons (OV>/=CV)
THIS IS A PARAMETRIC TEST
Describe the different levels of data
Nomianal: data in categories/groups
Ordinal: data on a scale
Ratio: actual numbers
Describe the criteria for a parametric test
If you’ve decided to go with a parametric test (unrelated t-test, related t-test or Pearsons) then you must fulfil 3 criteria:
- interval/ratio data
- homogeneity of variance: the standard deviation of one condition can’t be more than double the other
- data must be drawn from a normally distributed population; i.e. the mean, median and mode must all be the same
If any of these criteria aren’t met, you have to jump up the ladder once on your stats table, which we can’t get on here :( i.e. if you WERE going to do a Pearsons test but the criteria weren’t met then you’d jump up one and do a Spearmans rho etc
There are no special rules for the non-parametric tests
Describe how we test for significance
Each test has its own rule; either the OV is bigger than or equal to the CV -big wah- (Chi-squared, Spearmans rho, Unrelated t-test, Related t-test or Pearsons) or OV is smaller than or equal to the CV -little wah- (Sign test, Mann-Whitney U or Wilcoxon). If the rule is met then it’s significant and we accept H1 and reject H0
Sometimes we have to work out degrees of freedom…
- Unrelated t-test: df = N1 + N2 -2
- Related t-test: df = N - 1
- Chi-squared: df = (rows - 1)(columns - 1) on a contingency table
Degrees of freedom are used instead of just the plain no of ppts; N = no of ppts overall, N1 = no of ppts in C1 etc
Describe and explain the levels of significance we take
Typically use 5% (p= 0.05) i.e. 95% sure that the results are due to the manipulation of the IV, but there’s a 5% chance they’re due to other chance factors. Use the 1% level for drugs i.e. p= 0.01
5% level is used to avoid type I or II errors…
- Type I error: “false positive” i.e. accepting H1 when we actually shouldn’t have so claiming sig findings when they’re actually insignificant. Done from using too relaxed of a sig level e.g. 10% (p= 0.1)
- Type II error: “false negative” i.e. accepting H0 when we actually shouldn’t have so claiming insignificant findings when they actually are significant. Done from using too strict of a sig level e.g. 1% (p=0.01)
Describe the 3 pieces of info we need to use a critical values table
1) Is the hypothesis one tailed or two tailed?
2) No. of ppts in each condition (or degrees of freedom)
3) Level of significance being used