Scribd
Upload
5 views3 pages

Multivariate Stats Notes

The document provides an overview of key concepts in regression analysis, including R², assumptions of multivariate regression, and the differences between linear and multiple regression. It also discusses techniques like PCA, LDA, and clustering methods, highlighting their purposes and limitations. Additionally, it addresses issues like heteroscedasticity, multicollinearity, and the curse of dimensionality.

Uploaded by

Danyal
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views3 pages

Multivariate Stats Notes

The document provides an overview of key concepts in regression analysis, including R², assumptions of multivariate regression, and the differences between linear and multiple regression. It also discusses techniques like PCA, LDA, and clustering methods, highlighting their purposes and limitations. Additionally, it addresses issues like heteroscedasticity, multicollinearity, and the curse of dimensionality.

Uploaded by

Danyal
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Q: What is R² in regression?

A: It measures how much variation in the dependent variable is explained by the model (0 to 1).

Q: Heteroscedasticity & multicollinearity?

A: Heteroscedasticity: unequal variance of errors. Multicollinearity: predictors are highly correlated.

Q: Assumptions of multivariate regression?

A: Linearity, independence, normality, no multicollinearity, homoscedasticity.

Q: Methods in discriminant analysis?

A: LDA, QDA.

Q: Drawback of PCA?

A: Loss of interpretability and ignores class labels.

Q: Latent variable?

A: Hidden variable inferred from observed data.

Q: Cluster assignment example?

A: Distance: |35-15|=20, |35-30|=5 → assign to cluster C2.

Q: Shape of Gaussian?

A: Bell-shaped.

Q: What makes regression multiple?

A: More than one independent variable.

Q: Need of MANCOVA?

A: Compare multiple dependent variables while controlling covariates.

Q: Why LDA?

A: For supervised dimensionality reduction; assumes normal distribution.

Q: Is normalization needed for PCA?

A: Yes, for fair scaling.

Q: Outliers using IQR?

A: Values outside Q1-1.5IQR or Q3+1.5IQR.

Q: Linear vs Multiple regression?

A: Linear: one predictor; Multiple: many predictors.

Q: MANOVA/MANCOVA?
A: MANOVA: multiple outputs; MANCOVA: includes covariates.

Q: Curse of dimensionality?

A: High dimensions reduce model performance due to sparsity.

Q: K-means clustering?

A: Groups data into k clusters based on distance to centroids.

Q: Purpose of multivariate regression?

A: Predict multiple dependent variables.

Q: Difference multivariate vs multiple?

A: Multivariate: many outputs; Multiple: many inputs.

Q: ANOVA vs MANOVA vs ANCOVA?

A: ANOVA: one output; MANOVA: multiple outputs; ANCOVA: includes covariates.

Q: Job of discriminant analysis?

A: Reduce dimensions and classify data.

Q: Uses of LDA?

A: Classification, dimensionality reduction.

Q: First principal component?

A: Direction of maximum variance.

Q: Rotation in PCA?

A: Improves interpretability.

Q: Nonlinear data in PCA?

A: Not effective; PCA is linear.

Q: Limitations of PCA?

A: Linear, ignores labels, sensitive to scaling.

Q: Factor analysis?

A: Find hidden factors from observed variables.

Q: Effect of factor analysis?

A: Simplifies data structure.

Q: Limitations of factor analysis?


A: Subjective interpretation, needs large data.

Q: Remove correlated variables before PCA?

A: Not necessary; PCA handles correlation.

Q: Evaluate dimensionality reduction?

A: Use explained variance, reconstruction error.

Q: Why dimensionality reduction?

A: Reduce complexity, noise; drawback: info loss.

You might also like