Hello,
I am conducting a study where I measure thickness of tissue by analysing images. To do so I use calipers and I have to do it manually.
I need to obtain a variability extent of my measurements.
To do so I have measured the same tissue 10 times and need to find variability of my measurements to have an acceptable ability to detect differences.
What is the most acceptable method? Would SD and RSD suffice?
There are various alternatives to SD, like IQR or geometric SD (i.e., sd of the logs then transformed back into the correct units ... like the geometric mean), so it depends a bit what your measurement error is like, but like David Eugene Booth I would imagine your errors are fairly symmetric around the mean. Are they? If you have a few that are really way off you may want to consider alternatives. What will you be using this statistic for?
I would think sd would do. Be sure to include this in your statistical analysis. this means that if you use ANOVA (i.e. a linear statistical model so that includes regression) you need to look at an errors in variables (i.e. IVs) models. references are here: https://www.google.com/search?client=firefox-b-1-d&q=Errprs+in+Variables+Models If you decide not to use such a model be sure to mention that as a possible limitation of your study. Best wishes, D. Booth
There are various alternatives to SD, like IQR or geometric SD (i.e., sd of the logs then transformed back into the correct units ... like the geometric mean), so it depends a bit what your measurement error is like, but like David Eugene Booth I would imagine your errors are fairly symmetric around the mean. Are they? If you have a few that are really way off you may want to consider alternatives. What will you be using this statistic for?
Thank you both.
The way I have done it is this.
I took 25 spots of a scan and using a caliper I measured the thickness of each spot 10 times. I did this so as to have multiple random spots which I measured and also have more measurements in total to verify my method.
So in total I ended up with a 10x25 groups and a total of 250 measurements.
I then used the coefficient of variation or relative standard deviation as an attribute to tell me where my measurements stand, but as I had 25 groups of 10, I averaged the coefficient of variation.
Would this stand as a method?
@Daniel Wright
Yes if I remember correctly the 10 measurements of an individual spot usually came out at being symmetrical and a SD of 1. Normal distribution
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis