Hi,
I am trying to compare the results of two methods in a clinical laboratory. The data of the parallel measurements with the two methods as well as the differences between the measurements are clearly non-parametric (I have hundreds of observations; so n should be ok).
According to my Knowledge Bland-Altman Plot and Analysis assumes normality of the differences between the two methods, Deming Regression assumes normality of the data, while Passing-Bablok Regression assumes linear relationship between the two measurements, and in my data all these assumptions are violated.
I am thinking to use Spearman's Rank-Order Correlation, but I am not sure whether it really makes sense for evaluating laboratory method similarity, and I couldn't find a support to use it for my purpose in the literature.
I will be happy to read your suggestions!
Looks like you have already made all the test to see, that the methods are not interchangeable. No statictics can correct this fact. You have to do some other kind of validation (clinical) to see if the methods are useful.
Thanks Jukka for your answer,
I just made the basic tests in the guidelines for the verification of this assay!
Clinical validation is not possible for me.
The parametric procedure is also used when the population is not specifically the sample size possible to sufficiently large.
You can normalize data from a population by turning it into log. However, there are some suggestions for non-parametric testing..
Testes não paramétricos: Teste de sinal com 1 amostra: Teste de Wilcoxon para 1 amostra; Teste de Mann-Whitney; Teste de Kruskal-Wallis; Teste de mediana de Mood; este de Friedman
Thanks Paschoalina,
Log transformation didn't correct my data distribution. For me correlation is important; there are correlation tests for non parametric data, but I am not sure if they make sense in my case, as the correlation tests for non parametric data depend on the rank of the value and in clinical chemistry we are interested in the exact value.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression