Ravi_HMW_v4
Dihua Xu
May 22, 2017
Adiponectin versus Phenotype (0=Control, 1=Case)
##
## Welch Two Sample t-test
##
## data: Adiponectin, ng/ml by factor(Phenotype)
## t = 2.0083, df = 76.821, p-value = 0.04813
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## 16.82847 3964.77211
## sample estimates:
## mean in group 0 mean in group 1
## 10560.685 8569.885##
## Wilcoxon rank sum test with continuity correction
##
## data: Adiponectin, ng/ml by factor(Phenotype)
## W = 2291.5, p-value = 0.03742
## alternative hypothesis: true location shift is not equal to 0P values from both tests are < 0.05.
##
## Call:
## lm(formula = `Adiponectin, ng/ml` ~ factor(Phenotype) + `GA in weeks`,
## data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -8758.4 -4071.2 -906.5 3685.9 13487.6
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6271.1 3800.3 1.650 0.1014
## factor(Phenotype)1 -1849.1 979.6 -1.888 0.0614 .
## `GA in weeks` 387.7 339.8 1.141 0.2561
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 5191 on 128 degrees of freedom
## Multiple R-squared: 0.04119, Adjusted R-squared: 0.02621
## F-statistic: 2.749 on 2 and 128 DF, p-value: 0.06775##
## Call:
## lm(formula = `Adiponectin, ng/ml` ~ factor(Phenotype) + bmi,
## data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -10758.7 -3550.2 -530.8 3443.1 11720.5
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 19888.0 2633.1 7.553 7.09e-12 ***
## factor(Phenotype)1 -1221.4 954.3 -1.280 0.20292
## bmi -376.6 104.2 -3.615 0.00043 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4969 on 128 degrees of freedom
## Multiple R-squared: 0.1212, Adjusted R-squared: 0.1075
## F-statistic: 8.825 on 2 and 128 DF, p-value: 0.0002567##
## Call:
## lm(formula = `Adiponectin, ng/ml` ~ factor(Phenotype) + `GA in weeks` +
## bmi, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -10227.8 -3356.5 -556.6 3308.7 12537.1
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 15900.7 4520.9 3.517 0.000606 ***
## factor(Phenotype)1 -1099.3 960.3 -1.145 0.254454
## `GA in weeks` 352.8 325.3 1.085 0.280159
## bmi -373.2 104.1 -3.584 0.000481 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4966 on 127 degrees of freedom
## Multiple R-squared: 0.1293, Adjusted R-squared: 0.1087
## F-statistic: 6.284 on 3 and 127 DF, p-value: 0.0005205After adjusting for BMI and/or Gastational age, the Phenotype effect becomes non-significant.
## Area under the curve: 0.613
## threshold specificity sensitivity
## 1.028655e+04 4.943820e-01 7.380952e-01## dat$`Adiponectin, ng/ml` < 10287
## dat$Phenotype FALSE TRUE
## 0 44 45
## 1 11 31Area under the curve (AUC) for Adiponectin is 0.613. Best thresthold is “Adiponectin < 10287” as case (Phenotype = 1). The corresponding sensitivity is 0.738 and specificity is 0.494.
PEDF versus Phenotype (0=Control, 1=Case)
##
## Welch Two Sample t-test
##
## data: PEDF, ug/ml by factor(Phenotype)
## t = 0.50533, df = 70.048, p-value = 0.6149
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -1.657087 2.781787
## sample estimates:
## mean in group 0 mean in group 1
## 23.58957 23.02722##
## Wilcoxon rank sum test with continuity correction
##
## data: PEDF, ug/ml by factor(Phenotype)
## W = 1924, p-value = 0.7881
## alternative hypothesis: true location shift is not equal to 0P values from both tests are not significant.
##
## Call:
## lm(formula = `PEDF, ug/ml` ~ factor(Phenotype) + `GA in weeks`,
## data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -22.7008 -3.3717 -0.0519 3.9183 13.4345
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 22.65384 4.12349 5.494 2.04e-07 ***
## factor(Phenotype)1 -0.53144 1.06295 -0.500 0.618
## `GA in weeks` 0.08457 0.36874 0.229 0.819
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 5.632 on 128 degrees of freedom
## Multiple R-squared: 0.002626, Adjusted R-squared: -0.01296
## F-statistic: 0.1685 on 2 and 128 DF, p-value: 0.8451##
## Call:
## lm(formula = `PEDF, ug/ml` ~ factor(Phenotype) + bmi, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -23.8368 -3.5279 0.4128 3.6754 13.1650
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 15.3728 2.8913 5.317 4.55e-07 ***
## factor(Phenotype)1 -1.2402 1.0479 -1.183 0.23883
## bmi 0.3317 0.1144 2.901 0.00439 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 5.457 on 128 degrees of freedom
## Multiple R-squared: 0.06375, Adjusted R-squared: 0.04912
## F-statistic: 4.358 on 2 and 128 DF, p-value: 0.01476##
## Call:
## lm(formula = `PEDF, ug/ml` ~ factor(Phenotype) + `GA in weeks` +
## bmi, data = dat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -23.6551 -3.5699 0.3719 3.7730 13.0527
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 14.0652 4.9852 2.821 0.00555 **
## factor(Phenotype)1 -1.2001 1.0589 -1.133 0.25919
## `GA in weeks` 0.1157 0.3587 0.323 0.74757
## bmi 0.3328 0.1148 2.899 0.00442 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 5.476 on 127 degrees of freedom
## Multiple R-squared: 0.06452, Adjusted R-squared: 0.04242
## F-statistic: 2.92 on 3 and 127 DF, p-value: 0.03668P values are not significant.
## Area under the curve: 0.5147
## threshold specificity sensitivity
## 26.1258025 0.3707865 0.7142857## dat$`PEDF, ug/ml` < 26.125
## dat$Phenotype FALSE TRUE
## 0 33 56
## 1 12 30Area under the curve (AUC) for PEDF is 0.5147. Best thresthold is “PEDF < 26.12” as case (Phenotype = 1). The corresponding sensitivity is 0.714 and specificity is 0.371.
combined figure
## Area under the curve: 0.6212As we can see, the ROC curve of PEDF is almost always below Adiponectin. The combined AUC is 0.621.