Applying For History Jobs With Africana Studies PhD?

[Durchlässigkeitsbeiwert]

Another important piece of data

Interviewee suggests to "find multiple theses you can test with the same data".
Why it does remind me about somebody else p-hacking one's way to academic success?

Though, there are still some potentially useful conclusions to be made:
e.g. two out of 4 listed factors of academic success (the reputation of one's department, the reputation of one's adviser) are essentially set in stone once grad school starts. This implies that many grad students are more or less set to fail in their quest for "sweet gig" the moment they are admitted in.

[Caracal]

Another important piece of data

Interviewee suggests to "find multiple theses you can test with the same data".
Why it does remind me about somebody else p-hacking one's way to academic success?

Though, there are still some potentially useful conclusions to be made:
e.g. two out of 4 listed factors of academic success (the reputation of one's department, the reputation of one's adviser) are essentially set in stone once grad school starts. This implies that many grad students are more or less set to fail in their quest for "sweet gig" the moment they are admitted in.

Well, not really. Again, I would really encourage people to look for themselves. https://www.historians.org/wherehistorianswork

As you would expect, people from "stronger" programs do better, but I was expecting toes worse numbers from a lot of less prestigious programs. You can find a lot of not particularly elite state universities with over 50 percent of graduates in tenure track jobs. Even some regional universities are in the 30s. A lot of this is probably because even though there's a very competitive national market in the field, there are still a lot of regional and specialization effects that go into hiring. Getting a PHD from Kansas may not impress your uncle, but I'm sure they have some very good historians, probably especially in Western history. I'm guessing their graduates are heavily concentrated in jobs in the region.

[Hibush]

Another important piece of data

Interviewee suggests to "find multiple theses you can test with the same data".
Why it does remind me about somebody else p-hacking one's way to academic success?

He's not advocating p-hacking or data dredging. You  have to develop the thesis to test before you look at the data.

The advice is rather sound. If you develop a technique and set up your lab to run it well and have staff who are good at making it work, then it makes sense to ask a lot of different questions that can be answered with that technique.

Likewise, if you get an enormous genomics or public-survey data set, and have developed a really great R pipeline to ask a particular question, can you come up with other questions that can be tested with that same pipeline and data set? With those big data, the original question does not come anywhere close to exhausting the information contained. (Unlike us oldtimers who work hard for every datum and collect only those essential to doing the test we want.)

[Durchlässigkeitsbeiwert]

Interviewee suggests to "find multiple theses you can test with the same data".
Why it does remind me about somebody else p-hacking one's way to academic success?

He's not advocating p-hacking or data dredging. You  have to develop the thesis to test before you look at the data.

The advice is rather sound. If you develop a technique and set up your lab to run it well and have staff who are good at making it work, then it makes sense to ask a lot of different questions that can be answered with that technique.

Likewise, if you get an enormous genomics or public-survey data set, and have developed a really great R pipeline to ask a particular question, can you come up with other questions that can be tested with that same pipeline and data set? With those big data, the original question does not come anywhere close to exhausting the information contained. (Unlike us oldtimers who work hard for every datum and collect only those essential to doing the test we want.)

Developing 100 hypotheses before looking at the data, still makes many statistical tests meaningless, unless one starts to adjust thresholds used to determine statistical significance. Similar problem arises, if one applies the same scripted analysis routine to 100 sub-sets of a large dataset.