Should i use Multinomial Logistic Regression?

Hello to all.

I have an x-ray evaluation of patients. Each patient was seen in the clinic at different time intervals (Days between xrays), not regular times.
There are different rates of application to clinics for patients (at least 3, at most 10 or more times)
X-rays were evaluated for a certain difference, if an observer thinks there is a difference, he ticks "True", or vice versa.
All evaluation will be done by the same observers at 2 weeks interval, at a random fashion done by the software.

I am looking for:
1. How can we calculate the power of this study.
2. The lowest "Days between xrays" agreement between observers for all patients in general (interobserver & intraobserver correlation) (We may set "days" interval such as 100, 150, 200, 250, 300, etc)
3. Is there a difference of "lowest days between xrays" between sex?
4. Is there a difference of "lowest days between xrays" between different age margins? (30 - 50 years, 51 - 60 years, 60 - 70 years, 70 and over years)?

Thank you for your response.



Less is more. Stay pure. Stay poor.
I know you probably tried, but the description is difficult to interpret for the context and objectives. So what is the target outcome?

Do you not have a true status change variable for the patients (gold standard)? Are you just trying to find reliability between observers, if so how do you know who is truly correct. I cannot come up with any multinomial logistic reg scenarios given your description.

Perhaps it would help if you wrote out a model:

Radiological interpretation = Observer + Patient characteristics + ...

Though you have multiple radiologic images cluster within patients at non-uniformed time intervals? If you were writing a protocol what would be your primary aim, make it SMART by writing it then inserting variable names into it.
No, there is no gold standart. We are trying to find / establish a gold standart!
Target outcome is "There should be at least X days to observe radiological change on osteoarthritis, so if patient comes back for another visit earlier than X days, no need to study another x-ray of the same joint"

Non-uniformed time intervals, because this is not a prospective study, instead a database study, with x-rays studied earlier and present in the database.

Thank you