## DormanResearch : LearningStatisticsReferers: KarinDorman :: Learning :: (Remote :: Orphans :: Tree ) |
Dorman Wiki
Dorman Lab Wiki |

http://onlinestatbook.com/

Cross Validated via StackExchange

StatisticsHints

TimeSeries

FixedVsRandomEffects

NonParametricStatistics

AIC

LearningR

ClassificationMethods

Stat511

##### 0Regression

Cross Validated via StackExchange

StatisticsHints

TimeSeries

FixedVsRandomEffects

NonParametricStatistics

AIC

LearningR

ClassificationMethods

Stat511

- Ridge regression: method
- continuum regression
- Stone1990?
- merges PLSR and PCR on a continuum

- path analysis: used to test hypothesized cause/effect relationships
- regress each effect variable on its hypothesized causes
- compare regression weights against observed correlation matrix

- factor analysis: used to estimate latent structure under a set of variables
- use many variables to a few or confirm that one hidden factor explains many variables
- types: principal components analysis for data reduction, common factor analysis for causal analysis

- partial least squares regression (PLSR): like PCR, but identifies components of predictors in relation to response
- handles multiple responses, reducing both predictors and responses to principal components
- handles multicollinearity, missing values, and excess predictors
- useful for prediction, not so much explanation

- principal components regression (PCR):
- reduce potentially many predictors into the few most important principal components, which explain the most variation in the predictors.
- regression response on the principal components
- principal components are linear combinations of some orthogonal basis vector
- useful for prediction, not so much explanation, since principal components may be difficult to interpret

- structural equation modeling (SEM)
- models interactions, nonlinearities, correlated independents, correlated errors; basically a generalization of general linear model, including multiple regression, path analysis, factor analysis, and time series analysis
- largely confirmatory approach, cannot be used to infer structural model, except with repeated hypothesis testing and input from experts

- Nonlinear regression
- Fitting ODE models: http://www4.ncsu.edu/~ctk/Finkel_Direct/
- http://curvefit.com/index.htm

There is no comment on this page.
[Display comments/form]