For my master thesis I am conducting a research into Myasthenia Gravis (MG). The question I am trying to answer is: Does the AChR-concentration in a blood sample correlate with clinical symptoms (scaled by the MGFA)?
For this research I used a database with the outcomes of blood analyses in MG patients. I linked these samples to a score in clinical symptoms (MGFA). Next, I tried to pair a increase/decrease in concentration, with a increase/decrease in MGFA. However, some patients had more measurements than others. For example, one patient has 2 paired outcomes, and another patient has 8 paired outcomes.
Now I am searching for a legit statistical test to evaluate if there is a "correlation". I thought of using a Chi-square test for proportions. But I don't know if I can use this when there are multiple outcomes in one person.
Can somebody help me with this problem?
Daan Moens what is the distribution of MGFA? If the scores are a few ordinal categories, I suggest looking into cumulative link mixed models. If the scores follow Normal, maybe look into linear mixed models. The idea is that patient ID can be modeled as a random effect so that the relationship of MGFA and AChR-conc. isn't biased. What software platform are you using? I might be able to suggest packages/functions.
Caleb A. Aldridge MGFA has about 9 ordinal levels ranging from 0 till 5 (0-1-2a-2b-3a-3b-4a-4b-5). And the AChR-concentration is a normal continuous scale.
I am using SPSS 26 (IBM).
Side note: I am trying to simplify the statistical analysis as follows. For example, patient 1 has 5 measurements of paired MGFA and AChR-conc. I will then check if between every measurement the MGFA did go up, down or stay the same (3 possible outcomes) and if the AChR-conc. did go up or down (2 possible outcomes). This wil give me 4 interval measurements.
Patient 1:
T1: MGFA up, AChR up
T2: MGFA down, AChR down
T3: MGFA same, AChR up
T4: MGFA down, AChR down
I will do this for all other patients as well, but some have only 2 measurements (thus 1 interval) and others have 8 measurements (thus 7 intervals).
What I want to analyse is if there are significantly more paired measurements where MGFA did go up AND AChR did go up as well (and vice versa).
I hope you understand the question and my simplified method.
Since the same multiple measurements are recorded from each persons involved, this problem can be handled as a randomized block design. In that case the persons are the blocks. For it to be called randomized selection of the persons must be random. If all variables are measured on each person, then the design is called complete. Otherwise it is called incomplete. Then the analysis is done accordingly.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis