PSYC311 Flashcards
What is the dependent variable?
The numerical variable that measures what we are interested in.
What is an independent variable? What are they also called?
The condition that is manipulated by researchers to look for an effect. Also called factors.
What is between-subjects?
When subjects are exposed to one level of the factor.
What is within-subjects?
When subjects are exposed to all levels of the factors.
What is a mixed design?
When there are factors that are within-subjects and other factors that are between-subjects?
What is a pure within/between-subjects desgin?
When all factors in the study are all within/between-subjects.
What is a null hypothesis?
A prediction that the factor will have no effect on the population of interest.
What is a type I error?
When you reject the null hypothesis but should have accepted it.
What is a type II error?
When you support the null hypothesis but you should have rejected it.
If H0 was rejected, what two results could we end up getting?
A true positive (null is really false).
A false positive or type 1 error (null is actually correct).
If H0 was not rejected, what two results could we get?
A true negative (null is true).
A false negative or type II error (null should be rejected).
What does ANOVA test?
How the average of a dependent variable differs across categories.
What does correlation and regression analysis test?
How a dependent variable is related to another numerical value/s.
What does dummy variable regression and analysis of covariance (DVR & ANCOVA) test?
How does the relationship of one numerical DV to another numerical variables change across groups, conditions etc?
How do you find out the number of cells in the experiment?
Multiply the number of levels of all factors.
What are the 4 goals of the GLM?
- Representing the structure of the experiment.
- Measure the size of each factors effect.
- Measure the size of the sampling errors effect.
- Decide if factors are too large to be due to error.
What is the GLM model for a 1-factor between-ss design?
Y = baseline + A + S(A) Y = baseline + effect of factor A + error
What does the first letter in a Y subscript indicate?
Indicates the level of factor A.
What does the second letter in a Y subscript indicate?
The subject number for a 1 factor design or the level number for factor B
What does a dot in the subscript mean?
Average.
How to work out the baseline for estimation equations?
The average of all of the scores.
How to work out the baseline for estimation equations?
The average of all of the scores.
How to calculate the sums of squares.
Square every number in the column of decomposition matrix and add them together.
What does a large vs small sum of squares mean?
If there is a large sum of squares, there is a large effect of that variable. If there is a small sum of squares, the effect of that variable is small on the overall effect.
How to calculate df total.
Number of subjects x number of Y values per subject.
What is degrees of freedom for baseline?
1.
How to calculate df for any factor.
Number of levels of that factor - 1.
How to calculate the df for error.
Number of subjects - number of groups.
How do you calculate the mean square for a given term?
Sum of squares of that term / degrees of freedom of that term.
How do you calculate the F observed value?
Mean square of the given value / mean square of error.
When do we reject the null hypothesis in terms of F?
When the observed F value is larger than the critical F value.
What are the three equivalent versions of a null hypothesis? In english, in terms of the model, and in terms of comparison between models.
English: the true means are the same.
Model: the true population values of (insert term) are all zero.
Comparison: the model including (insert term) is just as good as the model without it.
What do we need to draw a causal conclusion? 3 steps.
Needs to be observed differences too large to be due to chance, subjects must be randomly assigned, and there must be no confounding variables.
What are we doing when observing a 2-way interaction?
Observing how the effects of A differ across levels of B (or vice versa).
State the model for a 2-factor between-ss design.
Yij = baseline + Ai + Bj + ABij + S(AB)ijk
How do we work out estimation equation for error?
Take the individual score and remove all other factors (baseline, factors, interactions). It is whatever is left over that has not been previously explained.
How to calculate the degrees of freedom for an interaction.
Multiply the degrees of freedom of all factors involved in the interaction (dfA x dfB x dfC).
Give a template for how to state an interaction effect between two factors.
The advantage of (insert two levels of one factor) is larger/smaller for (insert two levels of other factor).
What is the main thing to remember to show within an interpretation of an interaction?
Show how one factor differs over levels of another factor.
What happens if you ignore certain factors of the equation?
Error for each of the values is raised and we cannot get an accurate mean square or observed F.
What do unambiguous data patterns tell us? What are they?
When interaction effects go in opposite directions. Provides good evidence of dynamic interrelation for the effects of A and B, even with indirect (proxy) DVs.
What do somewhat ambiguous data patterns tell us about relationships between data? What are these patterns?
Interactions with an effect vs no effect. Suggests medium evidence for an interrelationship with proxy DVs, but strong evidence with direct DVs.
What do very ambiguous data patterns tell us about the relationship between factors? What is this pattern?
Interactions with small vs large effect and different starting levels. Provides weak evidence for interrelationship with proxy DVs, but somewhat strong evidence with direct DVs.
How can we tell from graphs if there is an interaction effect?
When the lines cross over, they diverge from one another, or there is more space between the lines at one end of the graph than the other.
What does a three-way interaction look at?
How a 2-way interaction changes across levels of the third factor.
How can we tell if there is a three-way interaction?
When the 2-way interactions look different in graph one compared to graph two (which shows the different levels of the third factor).
State the model for a 3-factor between-subjects design.
Yijkl = baseline + Ai + Bj + Ck + ACik + ABij + BCjk + ABCijk + S(ABC)ijkl.
What do the different subscripts in ‘Yijkl’ mean?
i = level of factor A. j = level of factor B. k = level of factor C. l = subject no.
How do you calculate two-factor marginal means?
Look at the means of two levels of certain two factors over the levels of the third factor.
How do you calculate one-factor marginal means?
Take the means for the individual levels of each factor.
How do we calculate the error terms in a shortcut for estimation equations? (For each individual score).
Take the individual cell score and subtract the cell mean.
