3

I have 10 different biological features (for example: normal, disease 1 , disease 2, ..., disease 9). Tissue samples from various patients were imaged and 60 imaging parameters were measured. A clinician identified different regions in the imaged tissue to be normal or diseased (1-9). I want to know if the measured parameters can distinguish between the different tissue conditions. I do not know if these parameters are independent of one another. There are thousands of values (for each parameter) assigned to a tissue condition. But this data was obtained by imaging a few tissue samples.

What sort of statistical test should I use to determine if the 60 parameters are statistically significantly different between the 10 different features? Is there anything in MATLAB? I was thinking of anovan.m, but it's not clear how I should go about using it.

hkf
  • 47
  • 5
  • I'd suggest that making clearer what you are doing would have a better chance of eliciting a good answer than offering a bounty. According to this you are estimating 600 parameters ... from what data? What kind of features? Examples and concrete details are needed. – Nick Cox Jun 11 '13 at 16:26
  • The most important question here is what your *n* is. What is a unit? Within each unit is disease status a polytomous variable (each unit takes one of either normal / disease 1 / ...) or a vector of binary indicators (e.g. a unit can be normal *and* have disease 2 *and* have disease 9). Are disease numbers ranked in severity, such as quartiles of viral load? Lastly describe the parameters: are they independent normal valued or discrete, is there measurement error, etc. – AdamO Jun 11 '13 at 16:34
  • yes it is a polytomous variable. The diseases are not ranked. They are different conditions observed by a clinician. I do not know if the parameters can be treated as independent of one another – hkf Jun 11 '13 at 17:02

1 Answers1

1

You want to look into multinomial logistic regression. This is available in MATLAB via mnrfit. There is an outstanding discussion of MLR on CV here: Interpreting exp(B) in multinomial logistic regression.

Community
  • 1
gung - Reinstate Monica
  • 132,789
  • 81
  • 357
  • 650