Intro to statistics Flashcards
What is a null hypothesis?
Difference between scores is so small that it is likely caused by chance.
What is an experimental/alternative hypothesis?
Any difference in scores between conditions is large enough that it is likely not caused by chance.
What is considered statistical significance?
1 occasion in 20, or P<0.05, or 5%.
What is a type 1 error?
Result is declared statistically significant where there is not a significant difference.
What is a type 2 error?
Result is declared not statistically significant where there is a significant difference.
What is a one-tailed hypothesis?
Predicts the direction of outcome.
What is a two-tailed hypothesis?
Does not predict the direction of outcome but still predicts statistical significance.
What is nominal data?
Data split into different categories with no order or direction e.g male/female, colours of sweets in a pack.
What is ordinal data?
Data has ordered categories where differences between them are not known e.g rankings/scales, places in a competition, you cannot assume the difference between 1st and 2nd place is the same as the difference between 3rd and 4th.
What is interval data?
Data measured on a numerical scale with equal distances between values e.g time, weight.
What are pairwise comparisons?
Comparing things in pairs to see if they are statistically different to each other.
What is effect size?
The magnitude of the relationship between variables.
What is the alpha criterion?
Probability of making a type 1 error.
What are post-hoc analyses?
Done after the data has been collected and analysed. Control for type 1 errors.
What are a-priori analyses?
Done before the experiment begins.
What is a chi-square stats test used for?
Analysing nominal data to compare the distribution of observations with those expected to be by chance.
When would you use a Yates correction?
When performing a chi-square test with 1 degree of freedom. Due to overestimation of the test statistic.
What is a T-test used for?
To compare means. Using interval or ratio data.
What is the standard error?
The standard amount by which a sample mean is in error when estimating the mean of the population.
What do T scores tell us?
In an independent they tell us how different our sample mean is from the population mean. In a related, they tell us the mean difference between two sets of scores.
What is a correlation?
A measure of the linear relationship between two variables.
When would a Pearson’s correlation be used?
For interval variables.
When would a Spearman’s Rho correlation be used?
For ordinal variables.
What is a correlation coefficient?
The amount of variance shared by the variables.
What do ANOVA’s tell us?
If there are statistical differences between independent groups. Then you can understand which independent variable(s) has a connection to your dependent variable.
What is a one-way ANOVA?
Compares the effects of an independent variable on one or multiple dependent variables.
What is a two (or more)-way ANOVA?
Compares the effects of more than one independent variable on one or multiple dependent variables.
What does a ‘factorial’ ANOVA mean?
Any ANOVA that has two or more independent categorical variables.
Why would ad-hoc tests be needed for an ANOVA?
As they do not tell you which groups are different from each other.
What are the assumptions of an ANOVA?
- Interval data
- Normality
- Homogeneity of variance
- Independence
What is the meaning of normality in stats?
Each ps supplies one piece of data and it is organized according to its numeric quantity, so the mean is accurate.
What is the meaning of homogeneity of variance?
The dependent variable scores have the same degree of variability across conditions
What is the meaning of independence in stats?
The absence of systematic bias from nuisance variables across and within conditions.
What are the methods of mediating the negative impacts of a repeated-measures experimental design?
- Counterbalancing
- Randomization
What is counterbalancing?
Half of participants do condition A then B, the other half do B then A.
What is randomization?
Different order of conditions is generated randomly for each participant.
Would using counterbalancing and randomization fix carryover effects?
No.
What are carryover effects?
The changes in participants that can be caused by a condition.
What is statistical power?
Probability of correctly rejecting a false null hypothesis, and finding an effect if it actually exists.
What is the minimum desired statistical power?
0.80 or an 80% chance of finding a true effect.
What can influence statistical power?
- Sensitivity of the study
- Size of effect of interest
- Alpha criterion
How can you know what statistical power to expect?
- Look at previous studies
- Run a pilot study
- Use Cohen’s ranges
How should power be calculated for factorial designs?
Required power should be calculated separately for all main effects and interactions.
When should non-parametric tests be used?
When assumptions for the parametric tests fail.
What is kurtosis?
The thinness of a distribution.
What are the 3 types of kurtosis?
leptokurtic = very thin/high to the top of the graph
mesokurtic = middle thinness
platykurtic = very fat/low to the bottom of the graph
What is the interquartile range?
three values that split the sorted data into 4 parts.
What are the quartiles?
Second = median
Lower = median of lower half of data
Upper = median of upper half of data
