PCA stands for Principle Principal Parametric Post-hoc( Principle, Principal, Parametric, Post-hoc ) Component Components Correlate Correlates Criterion Criteria( Component, Components, Correlate, Correlates, Criterion, Criteria ) Analysis Analyses Asshole Arsehole Array Arrays( Analysis, Analyses, Asshole, Arsehole, Array, Arrays ).
CCA stands for Canonical Correlate Correlation Correlates Correlations Component Components( Canonical, Correlate, Correlation, Correlates, Correlations, Component, Components ) Canonical Component Components Correlate Correlation Correlates Correlations Criterion Criteria( Canonical, Component, Components, Correlate, Correlation, Correlates, Correlations, Criterion, Criteria ) Analyses Analysis( Analyses, Analysis ).
The term representing the amount of original variance explained by a new derived variable is:
Eigenvalue
Eigenvector
Eigenvalues
Eigenvectors
The term representing weights showing how much each original variable contributes to each newly derived variable is:
The following information about PCA is True or False: First principal component – the vector on which the most data variation can be projected. Second principal component – vector perpendicular to the first, chosen so it contains as much of the remaining variation as possible.
The following information about PCA is True or False: First principal component – the vector on which the most data variation can be projected. Second principal component – Second best possible vector, chosen to account as much variation as possible, but less good fit than the First.
When to use PCA: You have a set of ‘p’ continuous class nominal explanatory( continuous, class, nominal, explanatory ) variables. You want to repackage their variance into ‘m’ components. You want ‘m’ to be < ≤ (<=) > ≥ (>=) ==( <, ≤ (<=), >, ≥ (>=), == ) ‘p’. Each component could/should/might explain different things.
Covariance or Correlation Matrix. If units of x and y are different, use a ❌ matrix (as it standardises the units). If units of x and y are the same (e.g. temperature) or with similar orders of magnitude, use a ❌ matrix (although you may need to standardise units).
In regards to the scree plot, a component with eigenvalue < 1 captured less than what?
1 variable’s worth of variance
1% of the total variance
1 average component's worth of variance
1% of the (1st) principal component's variance
Rotations, orthogonal vs oblique. Varimax is an example of Orthogonal Oblique( Orthogonal, Oblique ), meaning it does not does( does not, does ) allow for factors to correlate.
