Lecture 12 Flashcards
What is the main purpose of DFA?
To develop a classification rule to individuals to a correct group (IV), based on their scores of various predictors (DVs). For instance, a person could be assigned to the group of “hypokinetic dysarthria” based on their scores for jitter, shimmer and noise to harmonics ratio.
When should DFA be used?
When the grouping is already known
How is the classification rule developed?
- Each individual from Group 1 and Group 2 is measured on a set of predictors.
- The predictors are multiplied by coefficients, and then added together to form a z score:
z = a1x1 + a2x2 + a3x3 … - The coefficients are assigned such that the groups will maximally differ (and therefore there will be a more distinct difference between Group 1 and Group 2)
- The mean z scores for Group 1 (e.g. hypokinetic dysarthria) and Group 2 (e.g. normal voice) will be calculated, and future individuals will be classified into a group based on their z score.
What does Wilk’s Lambda tell us in DFA?
Wilk’s Lambda is a test statistic that explains the proportion of the total variance in discriminant scores NOT explained by differences between the groups (similar to the F ratio). Smaller values of Wilk’s are desirable.
There is a p value associated with this statistic. If it is significant, then we can reject the possibility that the level of prediction obtained could have been due to chance.
What is cluster analysis?
Cluster analysis is used to classify an originally unclassified group of participants on a set of dependent measures. It is an exploratory tool to identify possible groupings (homogenous clusters) from a set of DVs.
When are multivariate (as opposed to univariate) statistics used?
Multivariate statistics are applied whenever there is more than one dependent variable being analysed simultaneously.
They are contrasted with univariate statistics where only one DV is analysed at a time.
What is factor analysis?
In factor analysis, several correlated DVs are “reduced” or represented by a new single variable.
For example, an individual’s underlying intelligence relies on MANY components (DVs). However these have been “reduced” to two factors - verbal and non-verbal intelligence, both of which comprise several other measures (e.g. vocabulary, picture sorting).
