Tweet
Checking for hetereoscedascisity & proportional bias
Hi All,
I am currently assessing the validity of a new device, in order to check whether there is proportional bias I have examined the correlation between the residuals of the two measures (new device and gold standard) and the mean value measured from both devices. To assess if the data is heteroscedastic I have analysed the correlation between absolute residuals and the mean measured value - both returned low and non-significant correlations therefore i have assumed there is no proportional bias and the data is homoscedastic. Any advice as to whether I have done this wrong and how I could rectify this mistake would be greatly appreciated. Thank you in advance!
I am currently assessing the validity of a new device, in order to check whether there is proportional bias I have examined the correlation between the residuals of the two measures (new device and gold standard) and the mean value measured from both devices. To assess if the data is heteroscedastic I have analysed the correlation between absolute residuals and the mean measured value - both returned low and non-significant correlations therefore i have assumed there is no proportional bias and the data is homoscedastic. Any advice as to whether I have done this wrong and how I could rectify this mistake would be greatly appreciated. Thank you in advance!