Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- library(mosaic)
- # Einlesen in R
- tips <- read.csv2("tips.csv")
- mean(tip ~ 1, data = tips)
- lm(tip ~ 1, data = tips)
- #ALtdaten alte VL
- erglm1 <- lm(tip~total_bill, data = tips )
- erglm2 <- lm(tip~smoker , data=tips)
- summary(erglm1)
- summary(erglm2)
- #Neues
- erglm3 <- lm(tip ~ # abbhängige Variable
- total_bill + smoker, # unabhängige Variablen
- data = tips) # Datensatz
- summary(erglm3)
- plotModel(erglm3)
- #Bootstrapping
- set.seed(1896) # Reproduzierbarkeit
- Bootvtlg <- do(10000) * lm(tip ~ total_bill + smoker,
- data = resample(tips))
- confint(Bootvtlg)
- #Interatkionseffekt
- erglm4 <- lm(tip ~ total_bill + smoker + total_bill:smoker,
- data = tips)
- plotModel(erglm4)
- #Test alle Variablen
- lm(tip ~ .,
- data = tips)
- #Schrittweise Regression
- #Wertet Schrittweise alle Variablen aus und gibt aus, welche relevant sind
- #Noch ohne Interaktionen
- step(lm(tip ~ .,
- data = tips))
- step(lm(tip ~ .+total_bill:smoker,
- data = tips))
- #S. 311
- erglmtb <- lm(total_bill ~ size+time,
- data = tips)
- plotModel(erglmtb)
- #Logistische Regression
- gf_point( (sex=="Male") ~ total_bill,
- data = tips)
- #S.322
- # Referenzklasse festlegen
- tips$sex<- relevel(tips$sex, ref = "Female")
- # Kontrolle
- levels(tips$sex)
- # Speichere Ergebnis der Regression glm() in "ergglm1"
- ergglm1 <- glm(sex ~ # abhängige Variable
- total_bill, # unabghängige Variable(n)
- data = tips, # Datensatz
- # Abhängige Variable binomial,
- # Verknüpfung Logit
- family = binomial("logit"))
- summary(ergglm1)
- #Modell auflösen x = 10
- exp(-0.12 +0.04 * 10) / (1+exp(-0.12 +0.04 * 10))
- plotModel(ergglm1)
- set.seed(1896) # Reproduzierbarkeit
- Bootvtlg <- do(10000) *
- glm(sex ~ total_bill,
- data = resample(tips),
- family = binomial("logit"))
- gf_histogram( ~ total_bill, data = Bootvtlg)
- quantile( ~ total_bill, data = Bootvtlg,
- probs = c(0.025, 0.975))
- set.seed(1896) # Reproduzierbarkeit
- Nullvtlg <- do(10000) *
- glm(sex ~ shuffle(total_bill),
- data = tips,
- family = binomial("logit"))
- gf_histogram( ~ total_bill, data = Nullvtlg)
- prop( ~ abs(total_bill) >= coef(ergglm1)[2],
- data=Nullvtlg )
- # Ab hier VL 7 von 2019-11-19
- ergglm2 <- glm(sex ~ size,
- data = tips,
- family = binomial("logit"))
- summary(ergglm2)
- # S.334
- #Modell auflösen x = 10
- # b0 = 0,08 b1 = 0,2
- # 0,08 + 0,2 * x
- # x entweder 1, 2 oder 4
- 0.08+0.2
- # p = 0,28
- 0.08+0.2*2
- # p = 0,48
- 0.08+0.2*4
- # p 0,28
- exp(0.08 +0.2 * 1) / (1+exp(0.08 +0.2 * 1))
- exp(0.08 +0.2 * 2) / (1+exp(0.08 +0.2 * 2))
- exp(0.08 +0.2 * 4) / (1+exp(0.08 +0.2 * 4))
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement