Nếu chỉ so sánh 2 biến thì chỉ cần dùng so sánh 2 số trung bình
b= read.csv ("C:/Users/ADMIN/Desktop/SO LIEU TONG GD1.csv")
head (b)
## id type time PTs1 PTs2 PT.1 PT.2 PTr1 PTr2 aPTTs1 aPTTs2 aPTTr1 aPTTr2
## 1 2 1 0 11.5 11.7 100 97 1.00 1.02 37.7 38.0 1.16 1.17
## 2 3 1 0 11.9 11.7 95 97 1.04 1.02 38.2 38.9 1.17 1.19
## 3 10 1 0 12.0 11.9 93 95 1.04 1.04 38.4 38.4 1.18 1.18
## 4 29 1 0 12.0 11.8 93 96 1.04 1.03 39.9 39.8 1.22 1.22
## 5 36 1 0 12.0 12.0 93 93 1.04 1.04 38.8 38.6 1.19 1.18
## 6 49 1 0 12.0 11.5 93 100 1.04 1.00 38.2 38.2 1.17 1.17
## Fib1 Fib2 TT1 TT2 TTr1 TTr2
## 1 2.39 2.32 18.7 18.5 1.36 1.35
## 2 2.26 2.53 18.2 18.2 1.33 1.33
## 3 2.37 2.11 18.6 18.6 1.36 1.36
## 4 2.26 2.22 18.6 18.9 1.36 1.38
## 5 2.22 2.18 18.6 18.6 1.36 1.36
## 6 2.32 2.28 18.8 18.8 1.37 1.37
#Tính phương sai của PTs1 và PTs2
library (DescTools)
Desc (b$PTs1)
## -------------------------------------------------------------------------
## b$PTs1 (numeric)
##
## length n NAs unique 0s mean
## 3e+01 3e+01 0 1e+01 0 1.167e+01
## 100.0% 0.0% 0.0%
##
## .05 .10 .25 median .75 .90
## 1.110e+01 1.120e+01 1.128e+01 1.180e+01 1.200e+01 1.210e+01
##
## range sd vcoef mad IQR skew
## 1.300e+00 3.854e-01 3.302e-02 4.448e-01 7.250e-01 -4.016e-01
##
## meanCI
## 1.153e+01
## 1.181e+01
##
## .95
## 1.210e+01
##
## kurt
## -1.356e+00
##
##
## level freq perc cumfreq cumperc
## 1 10.9 1e+00 3.1% 1e+00 3.1%
## 2 11.1 2e+00 6.2% 3e+00 9.4%
## 3 11.2 5e+00 15.6% 8e+00 25.0%
## 4 11.3 2e+00 6.2% 1e+01 31.2%
## 5 11.5 2e+00 6.2% 1e+01 37.5%
## 6 11.6 1e+00 3.1% 1e+01 40.6%
## 7 11.7 1e+00 3.1% 1e+01 43.8%
## 8 11.8 4e+00 12.5% 2e+01 56.2%
## 9 11.9 4e+00 12.5% 2e+01 68.8%
## 10 12 4e+00 12.5% 3e+01 81.2%
## 11 12.1 5e+00 15.6% 3e+01 96.9%
## 12 12.2 1e+00 3.1% 3e+01 100.0%
Tính phương sai của PTs2
Desc (b$PTs2)
## -------------------------------------------------------------------------
## b$PTs2 (numeric)
##
## length n NAs unique 0s mean
## 3e+01 3e+01 0 1e+01 0 1.167e+01
## 100.0% 0.0% 0.0%
##
## .05 .10 .25 median .75 .90
## 1.130e+01 1.140e+01 1.147e+01 1.170e+01 1.190e+01 1.200e+01
##
## range sd vcoef mad IQR skew
## 1.200e+00 2.892e-01 2.479e-02 2.965e-01 4.250e-01 3.407e-01
##
## meanCI
## 1.156e+01
## 1.177e+01
##
## .95
## 1.214e+01
##
## kurt
## -7.487e-01
##
##
## level freq perc cumfreq cumperc
## 1 11.1 1e+00 3.1% 1e+00 3.1%
## 2 11.3 2e+00 6.2% 3e+00 9.4%
## 3 11.4 5e+00 15.6% 8e+00 25.0%
## 4 11.5 7e+00 21.9% 2e+01 46.9%
## 5 11.7 7e+00 21.9% 2e+01 68.8%
## 6 11.8 1e+00 3.1% 2e+01 71.9%
## 7 11.9 2e+00 6.2% 2e+01 78.1%
## 8 12 4e+00 12.5% 3e+01 90.6%
## 9 12.1 1e+00 3.1% 3e+01 93.8%
## 10 12.2 1e+00 3.1% 3e+01 96.9%
## 11 12.3 1e+00 3.1% 3e+01 100.0%
Tính phương sai giữa PTs1 và PTs2
lấy PTs1 trừ PTs2
length (b$PTs1); length (b$PTs2)
## [1] 32
## [1] 32
PT= b$PTs1 - b$PTs2
Desc (PT)
## -------------------------------------------------------------------------
## PT (numeric)
##
## length n NAs unique 0s mean meanCI
## 3e+01 3e+01 0 1e+01 3e+00 6.25e-03 -7.86e-02
## 100.0% 0.0% 9.4% 9.11e-02
##
## .05 .10 .25 median .75 .90 .95
## -3.00e-01 -2.90e-01 -2.00e-01 0.00 2.00e-01 2.90e-01 4.00e-01
##
## range sd vcoef mad IQR skew kurt
## 9.00e-01 2.35e-01 3.77e+01 2.97e-01 4.00e-01 2.15e-01 -1.05e+00
##
## lowest : -4.00e-01, -3.00e-01 (3e+00), -2.00e-01 (5e+00), -2.00e-01 (3e+00), -1.00e-01 (2e+00)
## highest: 2.00e-01 (3e+00), 2.00e-01 (3e+00), 3.00e-01, 4.00e-01 (2e+00), 5.00e-01
Kiểm định sự khác biệt giữa PTs1 và PTs2
wilcox.test(b$PTs1 , b$PTs2) #không thấy có sự khác biệt
## Warning in wilcox.test.default(b$PTs1, b$PTs2): cannot compute exact p-
## value with ties
##
## Wilcoxon rank sum test with continuity correction
##
## data: b$PTs1 and b$PTs2
## W = 538.5, p-value = 0.7257
## alternative hypothesis: true location shift is not equal to 0
#Nếu muốn dùng anova thì về kỹ thuật xử lý như sau
x1= c(b$aPTTs1, b$PTs2)
x2= c(rep('s1', length(b$PTs1)), rep('s2', length(b$PTs2)))
x= data.frame (x1, x2)
summary (aov (lm (x$x1 ~ x$x2)))
## Df Sum Sq Mean Sq F value Pr(>F)
## x$x2 1 8949 8949 1798 <2e-16 ***
## Residuals 62 309 5
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Tính sự khác biệt giữa PTs1 và PTs2
TukeyHSD(aov(x$x1 ~ x$x2))
## Tukey multiple comparisons of means
## 95% family-wise confidence level
##
## Fit: aov(formula = x$x1 ~ x$x2)
##
## $`x$x2`
## diff lwr upr p adj
## s2-s1 -23.65 -24.76492 -22.53508 0