Does anyone have software or procedures they have found useful for evaluating large, hard, inference problems? I don’t know what the right class of problem is. Mine is that I have several years and lots and lots of notes of symptoms a family member has exhibited, including subjective recollections all the way to MRIs, and I’d like to organize my thoughts and inferences around what common cause(s) might be, priors, weight of evidence, etc.
I think what GuySrinivasan’s asking is closer to “how do I organize a mass of evidence & ideas about a topic so I can better reason about it” than “how do I grind numerical statistical inferences out of a formal Bayesian model”?
One way to approach it would be to organize the data around the questions “What seems to have an effect on the system? What makes things better, what makes things worse, even if the effect is very small (but reproducible)?” Then, investigate those things.
Doctors are kind of terrible at doing that. They tend to have a tool box of “these are the things I know how to do” and any information that doesn’t fit their specific specialty is discarded as irrelevant.
I’m not sure how useful it would be to weight things by evidence if part of the problem is that some things haven’t been investigated enough, or are simply not well-enough understood by modern medicine and science.
I have a friend with an undiagnosed disease and am thinking about doing the same thing. One thing I’ve thought about is using a Bayesian Network as a tool, but then again, I’d have to be really careful about how I plug in data, and it would be good to know if there are other approaches to this as well. PM me if you find a good way to go about this.
Does anyone have software or procedures they have found useful for evaluating large, hard, inference problems? I don’t know what the right class of problem is. Mine is that I have several years and lots and lots of notes of symptoms a family member has exhibited, including subjective recollections all the way to MRIs, and I’d like to organize my thoughts and inferences around what common cause(s) might be, priors, weight of evidence, etc.
I plan to improvise, but I’d like to steal first.
Not sure what you mean.
BUGS maybe?
I think what GuySrinivasan’s asking is closer to “how do I organize a mass of evidence & ideas about a topic so I can better reason about it” than “how do I grind numerical statistical inferences out of a formal Bayesian model”?
One way to approach it would be to organize the data around the questions “What seems to have an effect on the system? What makes things better, what makes things worse, even if the effect is very small (but reproducible)?” Then, investigate those things.
Doctors are kind of terrible at doing that. They tend to have a tool box of “these are the things I know how to do” and any information that doesn’t fit their specific specialty is discarded as irrelevant.
I’m not sure how useful it would be to weight things by evidence if part of the problem is that some things haven’t been investigated enough, or are simply not well-enough understood by modern medicine and science.
I have a friend with an undiagnosed disease and am thinking about doing the same thing. One thing I’ve thought about is using a Bayesian Network as a tool, but then again, I’d have to be really careful about how I plug in data, and it would be good to know if there are other approaches to this as well. PM me if you find a good way to go about this.