Timeline for Proper analysis for vernal pool study data
Current License: CC BY-SA 4.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 16, 2023 at 16:26 | comment | added | DrKC | Obviously, if pH/conductivity is just reflective of how much water is in the pond, or the frogs do different things in identical ponds the math gets easier. | |
| Aug 16, 2023 at 16:19 | comment | added | DrKC | I think the biggest problem in making any recommendation about what direction to go is that we have no idea what happened. While I am asked to do so on a regular basis, I find making bold decisions w/ no idea of what the results of the tests show best left to politicians, as they don't seem to suffer from the same lack of certainty in the face of a paucity of evidence. Hopefully, we will find out what happened to the frogs in another installment. | |
| Aug 16, 2023 at 16:12 | comment | added | DrKC | I think what we know is there is an outcome variable of interest ( 4×9×3 measures of nascent tadpoles) and a data set which is thought to be predictive, but which really has two known variable (we don't know if the ponds are so similar as to make the between pond difference unimportant, I suspect the year to year variability may be much greater than between if they are separate ponds in a connected wetland ecosystem) but the measurements of pH and conductivity (which may, or may not be corrected to account for [H⁺]) both are numerators over the volume of water in the pond. | |
| Aug 16, 2023 at 16:08 | comment | added | Galen | Since you mentioned Bayesian networks, they can be used to model some forms of non-independence among observations. | |
| Aug 16, 2023 at 16:06 | comment | added | Galen | A Bayesian network is explicitly a probability distribution, representing a factorization of a joint distribution using the chain rule of probability, which relates to what I was saying about distributional assumptions being required to make inferences of statistical significance. Used in conjunction with KDE this makes the inference a nonparametric statistical procedure. There are no specifics to comment on unless they are provided, and without them there are too many possibilities to consider or discuss. | |
| Aug 16, 2023 at 15:53 | comment | added | DrKC | That doesn't have anything to do with the underlying problem when dealing with a small amount of information which is not independent and I don't have quite the statistical intuition to make any sort of recommendation without knowing what the data actually looks like. | |
| Aug 16, 2023 at 15:49 | comment | added | DrKC | I think the key is that ML models are not totally obscure. The value was the ability to pick out, from the mess of data that is a transplant patient's record, what was important and non-random. Again, it was about 12 years ago when he presented his work, so I am hoping that someone who works w/ various ML models--again, generalizations about "often" but rather looking for a particular ML--I want to say he was using a Bayesian network with kernel density estimation, but, as I mentioned, I was hoping to prompt someone who might be able to comment on the specifics, rather than just "most". | |
| S Aug 16, 2023 at 15:39 | review | First answers | |||
| Aug 16, 2023 at 16:26 | |||||
| S Aug 16, 2023 at 15:39 | history | edited | DrKC | CC BY-SA 4.0 | clairified variable use |
| Aug 16, 2023 at 15:34 | comment | added | Galen | [...] becomes infeasible to work out. And since we do not have IID assumptions, you cannot just bootstrap a p-value from retraining your ML model on resamplings of the data. Furthermore, ML models often have parameters which cannot be readily coupled to human-understandable hypothesis. | |
| Aug 16, 2023 at 15:34 | comment | added | Galen | "he basically fed his data into a few different machine learning frameworks and then [...] was able to pick out predictors and levels of significance." Leaving aside the problematic phrase "levels of significance", you will not obtain inferences of statistical significance without distributional assumptions. At that point, you might as well go back to classic statistical models since the change of variables for most ML models [...] | |
| S Aug 16, 2023 at 15:27 | review | First answers | |||
| Aug 16, 2023 at 15:36 | |||||
| S Aug 16, 2023 at 15:27 | history | answered | DrKC | CC BY-SA 4.0 |