Yes
No
To begin with, I am relatively unfamiliar with advanced statistics; I am familiar with elementary statistics such as Chi Square and Kruskal-Wallis tests used in my research. I have not, however, had the opportunity to take a formal statistics class yet. Therefore, I come to this forum with a question specific to my research...Thank you for understanding if the answer to my query is rather obvious to the experienced eye. With that formality out of the way
I cannot give too many details of my experiment in this medium. However, I am utilizing the staging of cancer disease, this particular one ranges from I to IV with subtypes, as the independent variable. I realize that this is ordinal or ranked. Then, I am figuring out the number of genetic components that fit each stage of disease. I want to see if a correlation exists between disease stage and said genetic component. For example, I would want to make a conclusion formatted like ,"As the stage of disease increases, the genetic component is increased (or as the results might indicate, decreased). How would I set up a correlation study in this scenario? Is it possible to set up a dummy variable in this instance?
Second conundrum. There are maybe x samples of stage x patient tumors and y samples of stage y patient tumors. Would a simple mean calculation account for this discrepancy or is there something more to do?
Again, thank you for reading my rather long, and first, post on this forum and I look forward to a solution to aforementioned problems.
Stop cowardice, ban guns!
What I mean is as follows:
There are 8 patient cases that are in stage IV and have 12 "genetic components".
There are 50 patient cases that are in stage IIIB and have 50 "genetic components" in total. (This of course is hypothetical data, but gives a good understanding of the level of differences). I considered stating that there are 12/8 or 1.5 "genetic components" that exist for each stage IV tumor sample and 1 "genetic component" per each stage IIIB tumor sample. Thus, I would plot (IV,1.5) and (IIIB, 1). Basically, I'm wondering if this way of measurement is valid, considering the fact that this is the only public data out there. Should I do some sort of a test to account for the vast differences in number of tumor samples. Hopefully that clarifies that.
What these genetic components I'm assessing for have to do with a specific kind of mutation type. So say it's a "green mutation", I would tally that for its respective tumor stage, and collect additional information on its role and interactions, etc. Hopefully that's enough information for you to assess my situation.
I didn't think that ordinal and continuous were synonymous. But I need a relatively simple (aka no heavy necessity for SPSS or related statistical software) for assessing a regression analysis.
Your guidance/input is much appreciated...due to the nature of my work it isn't prudent for me to disclose too many specific details but hopefully what I've provided will help.
Thanks,
NDDN
They are not synomynous. You would need the approach that is the most appropriate not simplistic, otherwise your results could be spurious!
Stop cowardice, ban guns!
Ok I looked up your suggestion of ordinal logistic regression and all youtube videos regarding the topic are relating the data to SPSS software which I have no way of obtaining. Is there a way of crunching the numbers without using SPSS?
I'm relatively good at hand crunching numbers--my math abilities are pretty strong.
Is there a simpler method...
What makes you think the method is only available in SPSS?
I don't have emotions and sometimes that makes me very sad.
Almost all methods that are available on a commerical software are going to be available on the other commerical software I would guess and R has even more if you can find out where.
I don't understand how your dependent variable is actually measured from the description, but ordinal varies from continuous in that the difference between levels of an ordinal variable are not the same while they are between a continous variable. In practice, if not theory, there are generally a small number of distinct ordinal levels while continuous has far more (in theory a continuous variable should have infinite levels although no real data base will do so). Commonly ordinal data will have five or fewer distinct levels. Some, although certainly not all, argue when you have enough distinct levels of a variable, say 11, then you can treat that variable as "interval like" effectively continuous even if the variable is actually ordinal.
"Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995
Tweet |