TL;DR- the difficulty of solving this problem is that availability of good data, NOT the lack of decent skill-set models.
So lets say you finish an outline of skills in these trees, as you suggest. Now, we want to make the statistical models that are at the core of your proposal- “One would determine what to learn based on statistical studies of what elements are, by and large, most desired by employers of/predictors of professional success in a certain field you want to work in.” Where exactly do you plan to get the data to actually do this? As you “level up” down your skill tree, the number of people who have skill combinations is going to get very, very thin. I might be the only person in the world who has both calculated a Next-to-Next-to-Leading-Order cross-section at Tevatron energy levels AND who has used restricted-Boltzmann-machines to categorize medical billing codes to triage health insurance claims. Backing up a bit, and “quantum field theory” + “health care informatics” + “statistics” might have what, a dozen people, all of whom are going to be outlier-ish for the general field of insurance data analysis.
Lets look at data from physics (because I know it well) here is the AIP’s focus on physics bachelor holders. Bachelors outnumber phds by quite a bit, so this is the densest you are going to get for certain types of physics skills. http://www.aip.org/statistics/trends/reports/empinibs0910.pdf. If you subtract out the exclusions (unemployed, people who went back to their pre-college job, and part-time employed), the data is already cut down to 2/3. Of those, the two largest chunks are highschool teachers (about 6% of the total physics bachelors) and engineers (about 7% of the total physics bachelors).
So we have about 350 people who teach highschool, probably highschool science, so you can maybe compare physics-educated general science teachers to other educated general science teachers. A bit more then that are in “engineering” fields, but there are tons of subfields/‘skill-trees’ in engineering. You’ll have <100 per skill-tree at the trees base! The data is already very thin, and its much thinner for all other career paths.
Career organizations that have much larger memberships (i.e. most career organizations) have much less detailed information because collecting it becomes more of a chore. These larger datasets probably won’t have the information you’d really need to fill out your skill-trees.
Also, I’m willing to bet that the world moves fast enough in a number of fields that you can’t use very deep historical data to do predictions. i.e. two decades ago a physics phds was a decent way to get a quant job on wallstreet, BUT two decades ago a masters of financial engineering degree didn’t exist. Maybe in another decade the MFEs squeeze out the phds the way that CS squeezed out physicist programmers in the 70s and 80s.
TL;DR- the difficulty of solving this problem is that availability of good data, NOT the lack of decent skill-set models.
So lets say you finish an outline of skills in these trees, as you suggest. Now, we want to make the statistical models that are at the core of your proposal- “One would determine what to learn based on statistical studies of what elements are, by and large, most desired by employers of/predictors of professional success in a certain field you want to work in.” Where exactly do you plan to get the data to actually do this? As you “level up” down your skill tree, the number of people who have skill combinations is going to get very, very thin. I might be the only person in the world who has both calculated a Next-to-Next-to-Leading-Order cross-section at Tevatron energy levels AND who has used restricted-Boltzmann-machines to categorize medical billing codes to triage health insurance claims. Backing up a bit, and “quantum field theory” + “health care informatics” + “statistics” might have what, a dozen people, all of whom are going to be outlier-ish for the general field of insurance data analysis.
Lets look at data from physics (because I know it well) here is the AIP’s focus on physics bachelor holders. Bachelors outnumber phds by quite a bit, so this is the densest you are going to get for certain types of physics skills. http://www.aip.org/statistics/trends/reports/empinibs0910.pdf. If you subtract out the exclusions (unemployed, people who went back to their pre-college job, and part-time employed), the data is already cut down to 2/3. Of those, the two largest chunks are highschool teachers (about 6% of the total physics bachelors) and engineers (about 7% of the total physics bachelors).
So we have about 350 people who teach highschool, probably highschool science, so you can maybe compare physics-educated general science teachers to other educated general science teachers. A bit more then that are in “engineering” fields, but there are tons of subfields/‘skill-trees’ in engineering. You’ll have <100 per skill-tree at the trees base! The data is already very thin, and its much thinner for all other career paths.
Career organizations that have much larger memberships (i.e. most career organizations) have much less detailed information because collecting it becomes more of a chore. These larger datasets probably won’t have the information you’d really need to fill out your skill-trees.
Also, I’m willing to bet that the world moves fast enough in a number of fields that you can’t use very deep historical data to do predictions. i.e. two decades ago a physics phds was a decent way to get a quant job on wallstreet, BUT two decades ago a masters of financial engineering degree didn’t exist. Maybe in another decade the MFEs squeeze out the phds the way that CS squeezed out physicist programmers in the 70s and 80s.