DanielEWeeks
commented
4 years ago I agree - in the function 'blandr.method.comparison', the x and y labels are switched in this line of the code:
plot(comparison.data$method2, comparison.data$method1, main = "Plot of two methods with line of equality",
xlab = "Method 1", ylab = "Method 2")
Apologies beforehand if my question isn't asked right.
In an attempt to understand a few things, I was looking at the
blandr.method.comparisonfunction and I noticed that method 1 and method 2 labels appear to be switched. I may be missing something, but can someone help me verify this? As shown below> function (method1, method2, sig.level = 0.95)
> {
> comparison.data <- blandr.data.preparation(method1, method2,
> sig.level)
> cat("\nNote as per Bland and Altman 1986 linear correlation DOES NOT mean agreement. ")
> cat("Data which seem to be in poor agreement can produce quite high correlations. ")
> cat("Line of equality in dashed black, linear regression model in solid red.\n")
> model <- lm(comparison.data$method1 ~ comparison.data$method2)
> multiplier <- model$coefficients[2]
> multiplier <- as.numeric(multiplier)
> intercept <- model$coefficients[1]
> intercept <- as.numeric(intercept)
> plot(comparison.data$method2, comparison.data$method1, main = "Plot of two methods with line of equality",
> xlab = "Method 1", ylab = "Method 2")
> abline(0, 1, lty = 2)
> abline(model, col = "red")
> results.paired.t.test <- t.test(comparison.data$method1,
> comparison.data$method2, paired = TRUE)
> cat("\nPaired T-tests evaluate for significant differences between the means of two sets of data. It does not test agreement, as the results of a T-test can be hidden by the distribution of differences. See the references for further reading.\n")
> cat("Paired T-test p-value: ", results.paired.t.test$p.value,
> "\n")
> cat("\nCorrelation coefficients only tell us the linear relationship between 2 variables and nothing about agreement.\n")
> cat("Correlation coefficient:", cor(comparison.data$method1,
> comparison.data$method2), "\n")
> cat("\nLinear regression models, are conceptually similar to correlation coefficients, and again tell us nothing about agreement.\n")
> cat("Using method 1 to predict the dependent method 2, using least squares regression.\n")
> cat("Regression equation: method 2 = ", multiplier, "x method 1 + ",
> intercept, "\n")
> }`
Since the plot function is
plot(x,y)shouldn't xlab then be method 2 and ylab be method 1?