library(tidyverse)
library(car)
DATE <- c("324", "325", "326", "327", "328", "329", "330", "331", "401",
          "402", "403", "404", "405", "406", "407", "408", "409", "410", "411", "412")
Domestic <- c(15, 14, 21, 83, 34, 33, 56, 87, 104, 160, 183, 133, 216, 281, 382, 384, 442, 431, 439, 551)
Temperature <- c(15, 19, 17, 16, 15, 18, 21, 17, 15, 13, 13, 17, 19, 18, 20, 20, 20, 21, 23, 22)

#氣溫以台灣首都”台北市的最低氣溫”為主，因為有流行病學專家說病毒比較容易在秋冬的溫度存活

建立一個資料名為cvdta

cvdta <- data.frame(DATE, Domestic, Temperature)

將DATE命名成一個類別變項

cvdta$DATE <- as.factor(cvdta$DATE)

用scatterplot圖檢視Domestic可不可以被Temperature預測及其關係，還有Domestic、Temperature分別的變異量

scatterplot(Domestic ~ Temperature, 
            data = cvdta, 
            smooth = F)

#可看出兩者關係呈現正相關

用ggplot圖繼續檢視兩者的線性關係

ggplot(aes(x = Temperature, y = Domestic), data = cvdta) +
  geom_point() +
  geom_smooth(method = lm, se = T) +
  theme_bw()

此圖是根據最小平方法繪製而成，灰色範圍越小越準確，標準差越小，表示估算越接近母體

將資料丟進簡單迴歸中

vmod <- lm(Domestic ~ Temperature, data = cvdta)
summary(vmod)
## 
## Call:
## lm(formula = Domestic ~ Temperature, data = cvdta)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -265.524  -73.800   -5.361  103.506  190.435 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)   
## (Intercept)  -498.33     200.99  -2.479   0.0233 * 
## Temperature    39.04      11.06   3.529   0.0024 **
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 138.6 on 18 degrees of freedom
## Multiple R-squared:  0.4089, Adjusted R-squared:  0.3761 
## F-statistic: 12.45 on 1 and 18 DF,  p-value: 0.002398

得出公式 : Domestic=39.04*Temperature-498.33

###結論

因為t value=3.592 >1.96，顯著大於0 ;p-value 0.0024 <0.05，所以可以拒絕虛無假設。

因為F-statistic=12.45 >1 ; p-value0.002398 <0.05，所以結果顯著

決定係數為0.3761表示依變項Domestic確診數可被自變項Temperature預測的變異量百分比有37.61%，高於20%，表示預測力佳