Goal
I have some data for Sublingual Microscopy (quantifies blood vessel stuff from under your tongue). I have X amount of patients split into two categorical groups (A & B). Each patient has 3 timepoints between measurements. Each individual timepoint has 2-6 individual measurements with its own data.
I want to see if the data for the blood vessels changes over time differently between group A & group B. I could, for each patient, average all 2-6 measurements per timepoint to create a dataset with only one value for each timepoint per patient (Then, run a one-way repeated measures ANOVA). BUT, I am concerned that some patient's 2-6 measurements might contain significant variance and thus be misrepresented by an average.
Question
Is there a test I can run to determine the variance for each patient & timepoint is small enough that an average would be representative for use in a one-way repeated measures ANOVA? I'll ask the same question in another way... How can I determine if my physical measuring tool is accurately measuring the patient?
For example, if I was testing out a new Heart Rate Reader, and sat a patient down and measured their Heart Rate 4 times in a row and got the values 30, 120, 65, & 10 I would say, "My Heart Rate Reader did not work! Maybe I used it wrong, or the patient was moving, or it was low battery. Regardless this data is not representative of my subjects Heart Rate in that moment" BUT, if a month later I took the SAME patient and got the output: 70, 71, 69, 72 I would say, "That worked! This data is representative of my subjects Heart Rate at this time because there is low variance."
Is there a way to objectively measure the lack of variance without bringing in the bias of human observation of saying, "Eh, looks good." or "I don't think that looks good."
THANK YOU!
Please see the above.