APA_Validation: GOF Im Folgenden die wichtige Voraussetzungen der...

In stp4/stp25stat: Functions for statistical computations

Description

Das Modell ist linear in den Parametern. Die Residuen sind normalverteilt und im Mittel sind die Residuen null. Die Zahl der Beobachtungen muss größer sein als die Zahl der zu schätzenden Parameter: n > k.

APA_Validation Validation of Linear Models Assumptions (Testing Linear Regression Models)

Multi-collinearity diagnostics

Conducts a series of checks for multicollinearity.

Bei perfekter Multikollinearität lässt sich das Modell nicht lösen.

APA_Autocorrelation: Der Durbi-Watson (DW) Test verwendet die geschaetzten Residuen um auf Autokorrelation erster Ordnung zu testen.

Der DW d Test weist folgendes Intervall auf: 0<d<4 0: Extrem positive Autokorrelation 4: Extrem negative Autokorrelation. 2: Keine Autokorrelation

Der Breusch-Godfrey (BG) Test ist flexibler, da er auf Autokorrelation hoeherer Ordnung testet.

APA_Heteroscedasticity: mit (lmtest::bptest) Breusch-Pagan test against heteroskedasticity

APA_Durbin_Watson: APA_Durbin_Watson(fit, max.lag=1, simulate=TRUE, reps=1000, method=c("resample","normal"), alternative=c("two.sided", "positive", "negative")): Durbin-Watson Test for Autocorrelated Errors. Kopie der Funktion car::durbinWatsonTest

APA2.durbinWatsonTest: Methode für car::durbinWatsonTest Kopie von car:::print.durbinWatsonTest

multicollinearity diagnostic information

VIF (variance inflation factor)

values over 5 are troubling, should probably investigate anything over 2.5.

Fuer Squareds und Correlations design matrix rockchalk::mcDiagnose(fit)

VIF.default lm Methode aus car::vif

APA_vif: variance inflation factor, aka VIF

VIF.merModLmerTest Quelle https://hlplab.wordpress.com/2011/02/24/diagnosing-collinearity-in-lme4/

Usage

APA_Validation(
  ...,
  include.ftest = TRUE,
  include.loglik = FALSE,
  include.minus.LL = include.loglik,
  include.pseudo = TRUE,
  include.r = include.pseudo,
  include.heteroskedasticity = TRUE,
  include.durbin = TRUE,
  include.levene = FALSE,
  include.bartlett = FALSE,
  include.multicollin = FALSE,
  include.vif = FALSE,
  include.sigma = FALSE,
  include.rmse = FALSE,
  include.aic = TRUE,
  include.bic = include.aic,
  include.residual = TRUE,
  include.normality = TRUE,
  include.deviance = TRUE,
  caption = "Testing Regression Models",
  note = "",
  names = NULL,
  output = which_output()
)

APA_Multicollinearity(
  x,
  caption = "Test for Multicollinearity ",
  note = "",
  include.vif = TRUE,
  include.cor = TRUE,
  output = which_output()
)

APA_Autocorrelation(x, order = 1, ...)

APA_Heteroscedasticity(x...)

APA_Durbin_Watson(
  x,
  caption = "Durbin-Watson Test for Autocorrelated Errors",
  note = NULL,
  ...
)

## S3 method for class 'durbinWatsonTest'
APA2(
  x,
  caption = "Durbin-Watson Test for Autocorrelated Errors",
  note = NULL,
  ...
)

VIF(x, ...)

## Default S3 method:
VIF(x, ...)

APA_vif(..., caption = "VIF", notes = "")

VIF2(..., caption = "VIF", notes = "")

## S3 method for class 'merModLmerTest'
VIF(x, ...)

Arguments

...	weitere Einstellungen
include.ftest, include.loglik	F-sratistik
include.r, include.pseudo	R-Quadrat
include.heteroskedasticity	Breusch-Pagan test
include.durbin	autocorrelation
include.levene	homogeneity of variance across groupssiehe T-Test
include.bartlett	Homogeneity of Variances siehe T-Test
include.vif	noch nicht Implementiert
include.sigma, include.rmse	RMSE Extract Residual Standard Deviation Sigma
caption, note	Ueberschrift an Output
x	fit-Objekt
include.cor	Bivariate Correlations

Details

• loglik Likelihood Ratio Test The log-likelihood from the intercept-only restricted model. -2LL: The LL (log-likelihood from the fitted model) llhNull (The log-likelihood from the intercept-only restricted model), G2 (Minus two times the difference in the log-likelihoods)

• Autocorrelation Durbin-Watson test for autocorrelation of disturbances. Ist nur bei Zeitreihendaten sinnvoll.

• Homogeneity of Variances Levene Computes Levene's test for homogeneity of variance across groups. Bartlett Test of Homogeneity of Variances

• Heteroskedasticity Breusch-Pagan test against heteroskedasticity.

• VIF variance inflation factor. VIF values over 5 are troubling, should probably investigate anything over 2.5.

• Residual RMSE values should be low (<0.5 and <0.3, respectively). SigmaResidual standard error When the residual standard error is exactly 0 then the model fits the data perfectly (likely due to overfitting)

• R-Quadrats Cox und Snell R2: [ 0.2 = akzeptabel, 0.4 = gut ] Nagelkerke R2: [ 0.2 = akzeptabel, 0.4 = gut, 0.5 = sehr gut] McFaddens R2: [ 0.2 = akzeptabel, 0.4 = gut ] (see pR2)

Examples

hkarz$Lai <- factor(hkarz$lai, 0:1, Cs(.neg, .pos))
fit2 <- glm(gruppe ~ tzell + Lai, hkarz, family = binomial)
fit1<-lm(score ~ grade + treatment + stdTest, schools)
APA_Validation(x) 
APA_Validation(fit1,fit2, include.pseudo = TRUE, include.r = TRUE, include.loglik = TRUE,
               include.rmse = TRUE)
               
               

# APA_Durbin_Watson
x<-lm(score ~ grade + treatment + stdTest, schools)
APA2(car::durbinWatsonTest(x))
DW_Test2(x)

lmtest::dwtest(x)
car::durbinWatsonTest(x)


## Not run: 
 fit<-lm(prestige ~ income + education, data=Duncan)
 car::vif(fit)
 VIF(fit)

## End(Not run)

stp4/stp25stat documentation built on Sept. 17, 2021, 2:03 p.m.