Inferential Statistics Flashcards
Statistical Tests
Used to establish if the results are significant or not. Were they due to chance?
Choosing a Test
- What level of data is used
- Is it measuring a difference or correlation
- What group design was used
Levels of Data
Nominal - data in categories. Data is frequencies and cannot be arranged in rank order and no scale of measurement.
Ordinal - data in rank orders. Can be put in rank order but no scale of measurement.
Interval - data is measured in units. Data can be measured on a scale that have equal units.
Hypotheses
They begin by writing a hypothesis. These are called alternative hypotheses as they are alternative to the null hypothesis (this states no difference). They will then either reject or accept this, and reject or accept the other one.
Probability
Measures the likeliness that an event will occur. Statistical tests work of the basis of probability.
Significance
Indicates degree of certainty that a difference or correlation exists.
Significance Level
They employ a significance level which is the point they can claim the results are significant or not. Usual level is p≤0.05 which means have an equal to or less then 5% chance occurred due to chance. They can never be 100% sure so settle at 5% due to chance.
Lowered Levels of Significance
Sometimes use more stringent levels in cases where human cost or can’t be replicated. This is to give more sureness that not due to chance. If larger difference between calculated and critical value in preferred direction will check with lower p value and if still within then lesser chance due to chance.
Errors
It is possible wrong hypothesis is accepted.
Type 1 Error = rejecting null when it is true and accepting alternative when it is false. False alarm. When significance level too relaxed.
Type 2 Error = accepting null when it is false and rejecting alterative when true. Missed detection. When significance level is too strict.
5% level balances between these.
