EPA MCQ Flashcards
What is the research process?
Identify variables, generate hypotheses, measure variables, analyse data
What is a sample?
A smaller but representative selection from the population used to make inferences about the population
What is the mean?
The sum of all scores divided by the number of scores, it is the value from which the scores deviate least (it has the least error) the mean could be represented as “Outcome + (Model) + Error”
How can we asses how well the mean represents reality?
Calculate error, the deviation between the mean and an actual data point.
What is the sum of squared errors?
It is when you find the deviation for each data point, square each vale (so they don’t cancel out by being +/-) and add them together
What is variance?
The sum of squares divided by the number of scores (to give an average)
What are degrees of freedom?
It is the maximum number of scores in a final calculation which can vary.
What is standard deviation?
The square root of variance. It shows how far away from the mean most data points are, it can be small or large around the same mean and make the graphs look different.
What is normally distributed data?
The mean, median and mode are all the same value, distribution is symmetrical around the mean, 99.7% of data falls within 3 standard deviations of the mean
What two tyles of skews are there?
Positive (towards the left) and negative (towards the right)
Two types of kurtosis?
Leptokurtic (pointy), platykurtic (round)
What is a Z score?
(Score-mean score)/ standard deviation. It is the umber of standard deviations away from the mean.
What are confidence intervals?
The boundaries in which we think the true population lies (this is different from simply the mean due to sampling variation)
What does a 95% confidence interval mean?
95% of the time the true mean will lie within the interval
What does a test statistic actually show?
Test statistic = variance explained by the model/variance not explained by the model = effect/error
What does the t-test show?
Difference in means(explained variance) / standard error (unexplained variance)
Type I vs type II errors
Type I = occurs when we think there is a genuine effect when in fact there isn’t (probability is alpha level) Type II = occurs when we thing there is no effect when in reality there is (probability is the beta level)
Issues with p-values
They are tied to sample size and so are not standardized.
What is an effect size?
A standardized measure of the size of an effect, the magnitude of the effect (e.g. eta2 etc.)
how do we determine if data is normally distributed?
Look at Zskew and Zkurtosis (statistic/standard error of statistic). If the score is +/-2 then the data is not normal.
What can you do with outliers?
Remove them. Transform the data.
What are the three types of data?
nominal (categories) Ordinal (ordered but with no true 0) continuous (numbers with a true 0)
Why can’t we just run lots of T-tests instead of an Anova?
Type I and II errors could be made. Due to the alpha level of .05
What is an anova?
An analysis of variance. Test statistic that lets us test if three or more sample means come from the same population.
What must be true to do an anova?
Data must be interval or ratio, sample distribution must be normal, there must be homogeneity of variance
What is homogeneity of variance?
Where the variance in each sample is equal (rather than one sample being really spread out and the others being almost identical)
Which test of homogeneity should I use?
Between design – Levenes, within design – Mauchly
Why do we have to do planned comparisons/post hocs with an anova?
Anovas produce an overall test statistic rather than specific information about each group/level of the data.
What do anovas actually do?
Compare the variance within groups to the variance between to show whether a difference in score sis due to the independent variable.
What type score do we want for a test of sphericity?
If the score is above .05 it means the assumption of homogeneity has not been violated, therefore we have homogeneity.
What corrections can be made to sphericity tests?
If greenhouse-geisser is smaller than .75 we make a G-G correction. If G-G is bigger than .75 we make a H-F correction. We make these corrections by reading different lines on the output from jamovi
How to report an anova result
F(DF for first variable, DF for second variable)=Fvalue, p=pvalue, eta squared or partial= value
How do we reduce chance of type I error in post hoc tests?
Use the Bonferroni correction - .05/number of comparisons
When do you use partial eta squared
Between groups is just eta squared, within groups is partial eta squared.
What are the non-parametric equivalents to an Anova?
Kruskal-Wallace for between groups, Friedmans for within/repeated measures.
What is a main effect?
The effect one independent variable has on the dependant variable ignoring any other effects. There can be multiple main effects!
What is an interaction?
When the effect of one IV changes across different levels of the other IV. When one variable has a different effect on different levels of the other IV. There can be multiple interactions!
What is the difference between post hocs and planned comparisons?
Post hocs are when we compare every condition with each other. Planned comparisons are when we only compare the variables needed to test our hypothesis
Why are planned comparisons ‘better’?
We only have to correct for a few comparisons rather than all of them ( so we don’t end up with an alpha level of like 0.000001)
What actually is the Bonferroni correction?
When we divide the alpha level (.05) by the number of comparisons we are running to account for familywise errors
What variables do we need to run an anova?
At least 1 independent variables and one dependent variable.
What is the non-parametric version of the multifactorial anova?
There isn’t one, we would remove outliers or transform the data.
How do we transform negatively skewed data?
Reflect the data (subtract the original value from a constant)and then square root the values.
How do we report the results of an anova?
State main effect (F(DF for first variable, DF for second variable)=Fvalue, p=pvalue, eta squared or partial= value), then interactions (F(df1,df2)= value, p= value, (partial) eta squared
What is statistical power?
The probability a study will detect an effect (when there is one) i.e. the probability of avoiding type II error.
How is statistical power expressed
1-beta
Limitations of null hypothesis significance testing
p-value is not standardised as it reflects the size of the sample (as well as the effect). So a small study can produce a non-sig result with the same effect size as a large study that is significant. It also encourages all or nothing thinking
if I have a significant Kruskal-Wallice test, what post hocs do I run?
Mann-Whitney
if I have a significant Friedmans test, what post hocs do I run?
Wilcoxon
From which lines of a results table do I get my two degrees of freedom values?
Corrected model (at the top) and error (above total)
What is the effect size of a Wilcoxon’s test ?
Z statistic divided by the square root of the number of observations
What does the Bonferroni correction do?
It corrects for familywise error
How many independent and dependent variables are there in a one-way repeated measures Anova?
Only one independent variable and only one dependent variable.
How many dependent variables must you have for an Anova to be conducted?
Only one at interval level or higher
How many levels must the independent variable have for an Anova to be conducted?
3 or more levels
What does Mauchly’s test do?
Indicates whether there is a significant difference between the variances of the conditions in a repeated measures Anova
What does partial eta squared measure?
Effect size
What does standard error measure?
The variability in scores in the sample
What is the relationship between sample size and the standard error pf the mean?
The standard error decreases as sample size increases
What does a 95% confidence interval mean?
95 out of 100 confidence intervals will contain the population mean.
Of what is P the probability?
The probability of observing a test statistic at least as big as the one we have if there were no effect in the population. The null hypothesis being true.
What is variance?
An estimate of average variability of a set of data
What is the non-parametric version of a one way between subjects ANOVA?
Kruskal-Wallice
What would you use for repeated measures instead of a one way Anova?
Friedman test
What is the test statistic based on for a Kruskal-Wallice test?
Ranks of the groups
What is the test statistic based on for a Friedmans test?
Sum of ranks
If the between groups variance is a lor larger than the within groups variance, the F value is…
Large
What is a large variance often due to?
A low sampling error
What is the overall effect of an independent variable known as?
Main effect
If a Mauchly’s test (conducted on a two way repeated measures Anova) has a value of p = .048 has the assumption of sphericity been met or violated?
Violated (as it is under .05 there is not sphericity)
If the Greenhouse-Geisser test has a value of .037 what correction should be applied?
Greenhouse-Geisser.
If the Greenhouse-Geisser test has a value of .080 what correction should be applied?
Huynh-Feldt
What is the primary use of a two way repeated measures Anova?
To assess the impact of two independent variables on a single dependent variable with repeated measures
In a two way repeated measures Anova what does the interaction effect test?
Whether the effect of one independent variable on the dependent variable depends on the level of the other independent variable
What is a key assumption of a two way repeated measures Anova?
There must be sphericity for both independent variables
In a two way repeated measures Anova what would a significant main effect for one of the independent variables mean?
That the dependent variable is significantly different across the levels of that independent variable, regardless of the other independent variable
What is the purpose of a Bonferroni correction?
It adjusts the significance threshold to control overall probability of making type one errors when performing multiple tests
What is statistical power?
The probability of rejecting the null hypothesis when it is false
What generally increases statistical power?
Increasing sample size
What does an a priori analysis help researchers determine?
The size of the sample needed to detect an effect
What is the relationship between effect size and statistical power?
As effect size increases, statistical power increases
