I have measured the same object 10 times in different areas and want to see if the 10 repeated measures agree. I have completed correlations which shows a stong rleationship but would like to look at agreement further. CanBland Altman be used in this case or would something else be better?
Take a look at the following
https://www.researchgate.net/post/How_to_manage_Bland_Altman_analysis_with_more_than_one_couple_of_measurements
https://www.researchgate.net/deref/http%3A%2F%2Fwww-users.york.ac.uk%2F~mb55%2Fmeas%2Fbland2007.pdf
There is not enough information to give useful focused advice. I'll give some general thoughts. Stating exactly what you need to answer would be a good place to start.
Have you samples got reproducible heterogeneity? Are the 10 different areas independently distinguishable and are they consistent across different samples? For example maybe the same type of tissue but in different organs?
Bland-Altman will be useful if area 1 in sample A is the direct equivalent of area 1 in sample B, while area 2 is distinct but is directly equivalent between each sample (and so on for all 10 areas).
If the samples are homogeneous or you cannot directly equate the same area number across samples then Bland Altman is not appropriate. You would want to explore sampling error to determine how many areas are needed to get a sampling error that is within reasonable limits (that depends on how sensitive you want to be to differences between samples).
Bland Altman eliminates the correlated element between sets of measurements expected to be equivalent so that you can examine the disagreement without distraction - i.e. you can focus on the disagreement or difference. This means each set of measurements needs to be directly matched with all the values in the other sets of measurements. In your example you will be testing if each area is exhibiting any bias (the mean of the values go through a value other than 0) and also the extent of the randomness between the sets of measurement (the scatter). The outcome will be
1) are the sites equivalent or are some biased?
2) how large is the random error between sites?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire