Posts tagged Department Data
People often ask me what the heck is going on with Cal Poly’s nutty quarter numbering system. I’m going to try to summarize it.
Before I begin, I think I’m probably suffering from Stockholm syndrome, here; I’m kind of fond of it, even though it’s non-intuitive.
Based on a very brief conversation with Debbie Dudley, it appears that the current system was designed in 2006, during the migration to PeopleSoft. At that time, Cal Poly had to devise a system for numbering quarters both in the future and in the past (that is, the new system also had to represent existing records of past students).
The rest of this is still largely hypothetical; please feel free to write and correct me about anything I’ve gotten wrong; I think that trying to understand how the system arose helps me (and you?) to understand its quirks a bit better.
(Semi) Hypothetical History
Problem: we need a way to represent a particular quarter (e.g.: Fall 2006) as a number. It would be really convenient to be able to sort in a sane way, etc. etc.
Solution: well, let’s use integers, not floats (for what are hopefully obvious reasons). Let’s use the year, then a digit for the quarter. What digits should we use for the quarters? Well, we should probably stay away from zero, because it’s not clear what year a zero would belong to. Okay, let’s use 2, 4, 6, and 8 for the four quarters. This spaces out the four quarters, and allows for the insertion of additional terms, should it become necessary.
So, how should we write Fall 2006? We could write “20068”. But… let’s try to cut that down to four digits.
(Knowing what I now know, I can’t see why you would want to do this unless PeopleSoft has some kind of internal restriction or performance penalty regarding numbers with more digits. Let’s assume this is the case.)
Well, maybe we can encode the century as a single digit. Let’s use “2” for the century from 2000 — 2099. So Fall 2006 becomes 2068.
Great! So how do we represent Spring 1997?
If we’re using century codes, then the one before “2” should be “1”. So in this case, Winter 1997 would be written as 1972.
Apparently, though, we decided that this was confusing, and that we should instead use a “0” for the 1900s, so that Winter 1997 is instead written as 0972.
This scheme is not awful, but it does have some problems if you try to subtract numbers or increment or decrement quarter numbers, because going from a fall quarter to the next winter quarter requires adding four… unless the year is 1999, in which case you have to add 1004. Oh well.
Okay, so let’s summarize. How the heck do we actually read one of these numbers?
Examples of quarter numbers:
2174 : 2000 + 17 = 2017, 4 = Spring. Spring 2017.
2102 : 2000 + 10 = 2010, 2 = Winter. Winter 2010.
976 : 1900 + 97 = 1997, 6 = Summer. Summer 1997.
Parsing a Cal Poly (San Luis Obispo) quarter number:
Actually, I’m going to be precise, and just give you some code.
Yes, in Racket.
;; given a quarter, return its season: "Fall", "Winter", etc.
(define (qtr->season [qtr : Natural]) : Season
(match (modulo qtr 10)
;; return the year in which a quarter number falls
(define (qtr->year [qtr : Natural]) : Natural
(define century-code (floor (/ qtr 1000)))
(define year-code (modulo (floor (/ qtr 10)) 100))
[_ (raise-argument-error 'qtr-year
"qtr with century code of 0 or 2"
(+ century-offset year-code))
Good news, folks! We’re continuing to deliver reasonably sized classes for our students. Specifically, more than 90% of our SCUs are currently delivered in classes of size less than 40. The last time we managed that was back in winter of 2011. On the other hand, if you think that class sizes should be below 30, we’re not doing so well.
Here’s the college as a whole (we’re better):
I’ve updated the dependency chart to match the 2017–2019 catalog.
This now includes DATA classes as well as CPE and CSC classes.
Some data: how long is it taking our students to graduate? How is that changing over time?
It looks like CSC and CPE are improving. It looks like CPE students are having trouble getting out in 4 years. On the other hand, the 5-year graduation rate for CPE majors is steadily rising.
It looks like lots of students graduate in four years and one quarter.
It looks like very few students graduate in their sixth year; that may be because there’s nobody left; this graph doesn’t distinguish between those who haven’t yet graduated, and those who are discontinued or disqualified. However, given the very small numbers for six years, it seems reasonable to assume that most of those that haven’t finished after six years already gave up, possibly years earlier.
Note that for students admitted in 2012, none of them are shown as having graduated in 4 years plus 2 quarters or more… because it hasn’t yet been that long since they were admitted. By the same token, we don’t see six-year numbers for 2011 admits, etc.
Also, all of this data uses “adjusted cohorts”. That is, those that change majors out of the program are removed from the denominator and the numerator, and those changing majors into the program are added to the denominator and the numerator.
(EDIT: images updated with 2174 data.)
Here’s a picture of (lecture, non-supervisory) sections taught by CSSE faculty since Fall 2008:
Because the number of sections taught has gone up rather dramatically, here’s the same data, normalized by sections per year, to show the fraction of classes taught in each category:
There are many caveats and things to explain about these pictures.
Following Zoe’s excellent suggestion, I added color to the course dependency chart:
The arrows should be the same as the previous chart, though you’ll notice that because of the way I generated it, the chart now includes the classes that don’t have any edges connected to them.
The colors relate to the number of students that have enrolled in each class. Specifically, the log of the number of students that have taken the class from fall 2010 through spring 2016. Classes with zero enrollment are given log(0.5). The class with the highest value is red, the one with the lowest value is green. Yes, there should be a legend. Classes that aren’t in the Computer Science catalog are gray.
Note that I have no idea how many of these enrolls are repeats. That would be interesting, but I’ll need a different data source to answer that question.
The numbers for each class are interesting, and I think I’ll just publish them separately.
(EDIT: updated with 2174 data.)
This picture shows the evolution of class sizes in classes taught by Computer Science faculty since Fall 2008. Specifically, it shows the likelihood that a randomly chosen scu will belong to a class of a given size.
So, for instance, you can see that in Fall 2008, more than 50% of our SCUs were delivered in classes of size less than 30, and that in Winter 2017 (2172), about 90% of our SCUs were delivered in classes of size 40 or less.
Here’s the picture for the College as a whole:
You may be wondering why I keep yammering on about SCUs, rather than saying things like “50% of classes…”.
Here’s another interesting picture. (Well, I thought it was interesting, anyway.) It shows the number of WTUs taught by the CS department faculty, from Fall 2008 up through Spring 2016. It includes courses with a bunch of different prefixes: CSC, CPE, HNRS, EE, ENGR, LAES, ME, and DATA.
This graph is broken up by the level of the course. The lowest (white) region shows courses whose names start with “01” (like “0123” and “0101”), the second region shows courses whose names start with “02”, and so forth. The “Sup” region shows the supervisory courses; senior project, master’s thesis, etc.
One note on “adjusted WTUs”: this data is taken from the FAD report, which misclassifies senior projects as lab courses, resulting in some very broken data. I’ve corrected this by re-assigning WTUs according to the CSU’s formulae.
Also, these are all classes taught by faculty associated with the department, so it includes lots of courses taught to nonmajors, as well as some courses taught by department faculty with other prefixes (for instance, a Mechanical Engineering course).
To me, the most interesting thing about this picture is frankly how flat it is. Our enrollments have gone way up, but the number of WTUs we’re teaching is pretty much unchanged.
I think the next picture to draw is how class sizes have changed over the years.
All parsing and rendering done in Racket. Isn’t it time that you learned Racket for yourself? :).
Here’s an SVG showing all of the dependencies associated with CSC courses:
Yes, it’s a little small to read. Click on it to see a bigger version. It’s an SVG, so you can blow it up arbitrarily. (Note: this picture is a lot more readable since Aaron Keen made the eminently sensible suggestion that it be left-to-right rather than bottom-to-top.)
Things to know about this data:
- It’s scraped from the 2015–2017 course catalog in HTML format.
- All cross-listed courses are normalized to their CSC equivalents.
- Arrows are shown to all courses mentioned in the prereqs.
The last of these is significant. If a course has a pre-requisite like “Both CSC 124 and one of MATH 117 or MATH 118”, I just draw arrows to all of them. So don’t assume that the number of outgoing arrows is an indication of the number of courses required to take this course.
- There are lots of courses shown here that haven’t been taught in a long time. CSC 108 jumps out at me, but there are others.
- Some courses have a prerequisite that can be fulfilled by a no-longer-existing course. For instance, CSC 141 changed into 348, but there are still a bunch of courses that list CSC 141. Since 141 is not displayed as a hyperlink in the catalog, we assume that it’s defunct, and we don’t show it.
- No dependencies are shown for non-CSC courses.
All scraping and processing done in Racket, natch. Graph drawn with Dot.
Okay, here’s a nice picture:
first-year retention pre- and post-change
This compares the outcomes in the years 2005 through 2009, before we instituted 123, with the outcomes in the years since. This is a cumulative graph, showing for a number of classes how many students finished by taking that many classes or fewer, and whether they succeeded or failed. Success is defined as having gotten a grade of C- or better in 103 (that is, the condition that allows them to continue with higher-level classes in the major).
So, for instance, we see that in the years before 123 was added, more than 90% of students took four classes or fewer, and that about 38% of those students did not successfully finish. In the years after 123 was added, slightly less than 90% of students took four classes or fewer, and of those, only about 23% left without finishing.
The big picture here is the post-change years have a much lower dropout rate; our first-year retention has improved from 63% to 77%, a dramatic increase.