L6 14/02 Flashcards
Nominal definition
Discrete categories or groups e.g., gender, age etc
Ordinal defintion
A pre-defined order e.g., very good, good, bad etc
Interval defintion
Where the distance between measures is the same, and where zero is arbitrary, e.g., years, percentage change
Ratio definition
Where the distance between measures is regular, but zero is absolute, e.g., a temperature scale
Features of normal distribution
- A symmetrical distribution (a bell-shaped curve)
- Measures are concentrated round the mean
- Decrease in frequency at each tail of the curve
- Suitable for PARAMETRIC tests – normal distribution
- Sigma represents standard deviations
Features of skewed distribution
- An asymmetrical distribution
- Positive (right) or negative (left) skew
- Positive is more common
- Suitable for NON-PARAMETRIC tests – skewed distribution
What are the 2 reasons tests are compared?
- Compare means
- Correlate data to determine trends
What are descriptive statistics used for?
Used to describe proportions
What are inferential statistics used for?
Used to estimate the likelihood that the results of a study are due to chance alone or that the sample results represent a true reflection of the population of interest. They allow us to measure the strength of a relationship.
What does statistical significance mean?
- A very specific term used to measure how confident we can be in the findings of research.
- Says that the results of your study are unlikely to have occurred by chance.
P=0.5 means
There is a 50:50 probability of the effect being due to chance (NOT a good result for a study)
P=0.05 means
There is a 1 in 20 probability (most studies use this for statistical significance)
P=0.01 means
Highly significant
What test would be most appropriate to compare the means of 2 normally distributed datasets?
T-test
What assumption do “parametric” statistics tests make about the data?
It is normally distributed
What does “statistical significance” mean about your results?
It is unlikely to have occurred due to chance
What does a p-value of 0.05 usually mean?
Statistical significance
What does a p-value of 0.01 usually mean?
Highly significant
For a positive correlation, what will happen to a variable if the other variable DECREASES in magnitude?
It will also decrease
What can “ANOVA” do that a “T-Test” cannot?
Compare 3 means
What is true about the Wilcoxon Mann-Whitney U Test?
It is non-parametric
P-value does not tell you what…
- It does not affect the strength of the effect:
- A highly significant result can be a tiny effect
- It just means that it probably does exist
- A high p-value does NOT mean that the effect does not exist; it just means you can’t prove it
