Thong ke suy dien

ATPSdata <- (AirTrafficPassengerStatistics) head(ATPSdata,3) str(ATPSdata) Newdata <- subset(ATPSdata, GEO.Region == 'Asia') head(Newdata,3) Newdata<- New_data[c("Operating.Airline.IATA.Code","Price.Category.Code","Activity.Type.Code","Passenger.Count","Year","Month")]

Hoi quy tuyen tinh da bien

Y=B0+B1.X1 + B2.X2 +....+ Bn.Xn + epsilon

Trong do Y la bien phu thuoc Adjusted.Passenger.Count

Tat ca cac bien con lai la bien doc lap

Mo hinh 1

Model1 <- lm(Passenger.Count~Operating.Airline.IATA.Code+Price.Category.Code+Activity.Type.Code+Year+Month,Newdata) summary(Model_1)

Kiem dinh cac he so Bi

H0: B1=0 (Y khong phu thuoc vao X1)

H1: B1 kh?c 0 (Y phu thuoc vao X1)

C?ch 1: p~value ( Pr(>|t|)) > muc y nghia 5%, chua bac bo H0 => B1=0 => X1 ko anh huong Y

p~value (Pr(>|t|))< muc y nghia 5%, chua bac bo H0 => B1=! 0 => X1 co anh huong Y

Cách 2: t~value: Thong ke kiem dinh

RR=(-vc, -talpha/2, n-k-1) U (talpha/2, n-k-1; +vc )=(-v; -1.961642) U (1.961642; +vc )

talpha/2, n-k-1 = t0.025, 1415 =

qt(p=0.025, df=3267, lower.tail=FALSE)# nguong 2 phia

Uoc luong (Estimate) | Sai so chuan (Std. Error)| Thong ke kiem dinh (t value) | Gia tri p (Pr(>|t|)) |

t value=Estimate/Std. Error

Noi dung: Kiem dinh duong hoi quy

H0: B1=B2=0 (R^2=0) => pt ko co y nghia (ko thich hop)

H1: ton tai Bi khac 0 => pt co y nghia ( thich hop)

F-statistic: 1684

RR= (Falpha;k;n-k-1; +vc)= (F0.05;5;3267;+vc)

qf(p=0.05, df1=5, df2= 3267, lower.tail=FALSE)

??y l? h??? s??? x?c d???nh (R?) c???a m? h?nh h???i quy tuy???n t?nh. N? do lu???ng t??? l??? bi???n thi?n c???a bi???n ph??? thu???c (AdjustedPassengerCount) du???c gi???i th?ch b???i c?c bi???n d???c l???p (Month v? Year) trong m? h?nh.

??y l? h??? s??? x?c d???nh di???u ch???nh (Adjusted R?). N? di???u ch???nh gi? tr??? R? d??? ph???n ?nh s??? lu???ng bi???n d???c l???p trong m? h?nh v? k?ch thu???c m???u. Adjusted R? thu???ng du???c s??? d???ng d??? so s?nh c?c m? h?nh v???i s??? lu???ng bi???n d???c l???p kh?c nhau.

(R?) thu???c kho???ng t??? 0 d???n 1

Residual standard error: sai s??? chu???n du???ng h???i quy ( l? sicma)

s=\sqrt(SSE/(n-k-1))

Residuals: sai s??? ng???u nhi?n, sai s??? h???i quy (epsilon_i=Y-Y^)

B??? Month,Price.Category.Code do Pr l???n hon 0.05

M? H?nh 2

Model2 <-lm(Passenger.Count~Operating.Airline.IATA.Code+Activity.Type.Code+Year,Newdata) summary(Model_2)

=> R^2 hi???u ch???nh tang l?n n?n m? h?nh t???t hon, ph? h???p hon

anova(Model1, Model2) par(mfrow=c(2,2)) plot(Model2) library(car) vif(Model2) X <- data.frame( Operating.Airline.IATA.Code = "CA", Activity.Type.Code = "Enplaned", Year = 2015 ) predict(Model_2, X, interval = "confidence", level = 0.98 )