In the coming weeks I hope to be updating you with more details and analyses, but for now I am simply announcing that the final report for APDA is complete. Feel free to ask questions or comment below.
*Update: we noticed an error in one of the charts and some potentially confusing language in that section, so we have updated the report at the link.
Thanks for sharing this, and all the hard work of data collection! I’ll probably have more questions when/if I have time to read through carefully. Here are a few quick ones:
* Pie charts, especially 3D pie charts, are often criticized for being difficult to read and potentially misleading (basically, lots of people are bad at estimating areas of circles, especially with foreshortening). I think stacked bar charts make more sense for the data you’re presenting in sections 3.3 – 3.5.
* For table 3, you’re testing against the null hypothesis that the coefficient = 1, correct? It took me a minute to figure that out. For non-stats-savvy readers it might be useful to explain how to interpret the estimates.
LikeLike
This may have come from an incomplete reading, but I’m not sure I understood the ratios in the report. Is 1.95 better than .68? For example, is it much better to seek employment in philosophy of science or history, in terms of the percentage of prior candidates who secured tenure track placement?
Also, when candidates had multiple areas of specialization, did they get binned in all of them? Again, sorry if a more detailed reading could have answered these, I was just wondering.
LikeLike
Hello Eli,
Thank you for your question. When interpreting an odds estimate, remember that it is a ratio between two groups. So ignoring the multilevel nesting component for a moment we could easily calculate the odds ratios by hand for gender, for example.
Pulling numbers from table 1 on page 12, the odds of males reporting a permanent position would be the number ‘yes’ divided by the total number of males: 269/681= .40
The odds of females reporting a permanent position would be calculated the same way: 138/257=.54
So females had better odds of reporting a permanent position. The ratio of females to males reporting a permanent position would then be: (.54)/(.40)=1.35
Therefore, female participants would be 1.35 times as likely as males to report a permanent position. The numbers in the report are slightly different, as some of the variability is accounted for by nesting within institutions, but can be interpreted in a similar manner. So, the Science, Logic and Mathematics are 1. 95 times more likely to report a permanent position as compared to the intercept, or Metaphysics and Epistemology. And History of Western Philosophy is 1.3 times as likely to report a permanent position as compared Metaphysics and Epistemology, however, this does not appear to be a statistically significant difference. Each group is not compared directly to on another, instead each is compared to Metaphysics and Epistemology. In a sense, Metaphysics is acting as a control group here.
LikeLike
I agree with your comments and for future updates will keep the figures and tables in a more standardized format. Also, you are correct in your interpretation of the null hypothesis. With logged dependent variables, we often test the null that the coefficients will be equal to 1. This is because the effect on the unlogged variable would equal 0. I’ve posted an explanation of estimate interpretations in the reply to the below question, as well.
Thank you!
LikeLike
Hi Eli,
Just to add–the multiple areas of specialization were not binned. We categorized only first reported AOS.
LikeLike
Hello again, I made a slight mistake in my previous response. The calculations should have been as such:
Males:
(odds of ‘yes’)/(odds of ‘no’)
(266/681)/(412/681) = .65
Females:
(odds of ‘yes’)/(odds of ‘no’)
(138/257)/(119/257) = 1.16
Thus, female respondents have better odds of reporting a permanent position. Their ratio would then be:
1.16/.65 = 1.78
Therefore, female participants would be 1.78 times as likely as males to report a permanent position.
I apologize for the error. Hopefully, that helps with clarification.
Best, Patrice
LikeLike
New APPS 
Leave a comment