Winter Weather

Oh hi! Didn't see you there.

It is now spring! And despite the massive continuous blizzard that appears to be going on outside, we're supposed to be getting warmer. Any day now...

You may have seen my analyses of Summer and Fall for Edmonton weather. Hopefully ever since then you've been on the edge of your seat awaiting the results for winter.

Wait no longer! The winner for winter is: The Weather Network. (three times in a row!)

Scores for winter (out of 100):

• Weather Network: 70.91
• Weather Channel:67.71
• Environment Canada:66.79
• Global Weather:65.45
• TimeandDate.com:60.05
• CTV Weather: 57.38

Noteworthy about these scores is that Environment Canada climbed from 5th place to 3rd place for the winter, and that everyone's scores (apart from Environment Canada's) continued to decrease from the fall. This is all shown in this graph:

Weather or not (PUN!) temperatures and precipitation are actually tougher to forecast in winter is a question better asked of the actual meteorologists. My suspicion is that at least part of the continued decrease in scores is that trace levels of snow are harder to measure as precipitation than rain, but that's mostly just a guess.

Some fun facts!

Best high temperature prediction: Weather Channel 1-day prediction: 71.84%
Best low temperature prediction: Weather Network 1-day prediction: 68.64%
Best precipitation prediction: Weather Network 1-day prediction: 76.67

Worst high temperature prediction: TimeandDate.com 6-day prediction: 36.30%
Worst low temperature prediction: TimeandDate.com 5-day prediction: 37.64%
Worst precipitation prediction: Environment Canada 6-day prediction: 54.70

Some graphs!

Again, CTV scores are only directly compared to the others for four days. I still find it cool that there is as strong of a downward trend as there is - on average, a forecast for a week in the future is 15% less accurate than a forecast for tomorrow.

For those of you who are still reading and like graphs, you can check out the breakdown of where the previous graph comes from:

Have a good spring!

Keep your hands off of my science

Science is great.

Say you want to see which medicine is the most effective at curing the flu. A good test would be to grab a group of sick people and give half of them treatment A and half of them treatment B, and see how they do.

There may be a couple problems with this, though. Maybe when picking the groups you do a bad job and get sicker people in one group than the other, or there's a noticeable age divide. A good way of countering the possibility of bias here is to have truly random group division. If you had a large enough group of people and tossed a coin on each to divide them into two groups, you could expect a reasonably fair trial.

What if the patients taking the medicine have heard rumors about treatment A or B, though? Maybe A seems more serious and they stress out about how ill they are, or they've heard that B is newly-developed and not proven? Fortunately, this is easily countered by doing a blind trial - don't tell any of the patients what medicine they're getting, and then compare the results.

But what if the doctors administering the medicine similarly have heard rumors about either treatment? Maybe they'd pay more attention to the patients with the perceived weaker treatment, or interpret the results to fit their expectations. Countering this has led to one of the pinnacles of scientific testing: the double-blind trial: neither the doctors nor the patients know who is getting what treatment. Only after all the testing is complete and the results are analyzed can the conclusions be actually drawn.

This pursuit of eliminating bias to get fair and true scientific results is one of the best features of science. The problem is that it doesn't stop there.

There is quite simply never enough funding for science. There are virtually unlimited questions about our world (about even just our own bodies) that have yet to be formulated, let alone answered, and there is only ever a limited amount of funding to cover all of the research to be done. Scientists clamor over each other trying to get the funding, which often comes either from government research centers or from corporations.

Now, I have nothing wrong with the idea of corporations investing money in research. I have a problem with how that can (sometimes deliberately) skew the results. While it may sound a bit like a conspiracy, study after study has shown that scientists know who pays the bills, and this has a significant impact on their results. These effects may not always be deliberate, but just like a patient may receive clues from a doctor on how well they think a treatment will work, a researcher may be aware that if they say a product is bad they won't get future funding from that company.

With that all said, I am extremely (bold, italics, and underlined) frustrated with the Government of Alberta's proposal to focus funding away from "curiosity-driven research" and align research funding to coincide with the province's "economic diversification agenda."

Presumably the government wants to make sure it's getting good value for its research dollar, but since when has curiosity-driven research been of low value? Newton and Galileo made great leaps in the name of curiosity, and the most sophisticated piece of technology outside of our world is called Curiosity. Some of the best things we know of like penicillin were only discovered by accident, and vaccinations were only discovered out of pure 'willing-to-sacrifice-little-kids' curiosity.

The only way to eliminate this last major obstacle of bias in science is to make the funding itself blinded. If a company anonymously sponsored scientists in another country to perform research on a drug that also remained anonymous to them, the rest of the research was double-blinded, and the results were made public, we would truly have the highest standard of scientific study. This is the exact opposite of the direction in which the government wants to take research.

At its worst, we could have a government who picks and chooses which cancers are important to study. Or which forms of mining should be improved. Or has an undue influence on the results of climate research. This is bad for innovation, creativity, and the production of true and objective science, and should be fiercely opposed.

Running for the SU?

Now that we're between the Students' Union executive elections and the Council elections, I'd like to take a second to clarify a bit a point I may not have successfully made during my post on the presidential platforms.

In very general terms, the portfolios of the SU Vice Presidents can be spectrumized (scientific term) like this:

On the left hand side we have portfolios that deal with issues external to the university, and on the right hand side we have portfolios that deal more with the running of the SU. The President would be expected to assist with and coordinate all of these, and balance their time accordingly.

An alternative way of looking at the internal/external label is, to quote Pirates of the Caribbean, "What a [duly elected executive] can do, and what a [successful student politician] can't do."

Before anyone gets upset at me, let me clarify - electoral promises that are made on the external side of things are still important. Being the elected representative for 30,000 undergraduate students is not insignificant. Sitting on the board of governors commands a fair amount of respect, and working with other student unions across the country on coordinated lobbying efforts is the most effective way to get the opinions of students heard.

The problem is that, at some (perhaps hyperbolically-simple) level, externally-oriented electoral promises are promises to bring things up and talk about them in meetings. Both candidates for president in the last election promised to research things or start dialogues. And though the SU's research team is phenomenal, and their arguments could be solid, fundamentally anything the SU brings up externally is subject to someone higher up just saying no.

On the other hand, as an executive of an organization, in control of the $10,000,000 budget, it is relatively straightforward to perform internal changes to the SU. Though it's not a great idea to have massive swings in the internal workings of the SU year after year, it is something that an exec would have the definitive say over, and as campaign promises they are much more tangible.

I mention all this also because a fun time is soon to be upon us: SU councillor elections. Yay! While exciting, it's important to keep in mind, once again, what you can do and what you can't do as a councillor when you're developing a platform.

If you're running for council to get the SU to use its considerable lobbying power for a pet project of yours, chances are it won't happen. At best you may be a member of the policy committee, which debates and votes on policy suggestions to council, where they're debated and voted on before being brought up in meetings with officials. On the other hand, if you want to have input on how the SU handles its advocacy, then an externally-oriented platform is legitimate.

More realistically, running for council can be a great opportunity instead to work on the internal portions of the SU. Maybe you hate/love APIRG and other dedicated fees students pay? Maybe you want an influence on the businesses the SU runs, like RATT and Dewey's? Maybe you want to help your Faculty Association regain its FAMF, get generally more involved, and overhaul the electoral system [LAME]?

I strongly encourage anyone even vaguely interested in running to run. The point of all of this is that there are plenty of actual important things you can run on, and a campaign based on external pet projects is likely to result in disappointment for all involved.

Good luck! And remember to get your nomination packages in before 5:00 on Tuesday...

SU Results Analysis 2013

This was a triumph. I'm making a note here: huge success.

Well, sort of. With only two races to check, having my model get 2/2 races correct could have been just chance. Let's take a look.

Yesterday my prediction was for Petros to beat Saadiq by a margin of 39.6%-35.1%, and William  to beat Kevin by a margin of 47.9%-35.5%. In reality, Petros won by a margin of 41.9%-38.8% and William beat Kevin by a margin of 53.0%-36.5%. On average, there was a 3.0% difference between what I predicted and what happened.

[Note: I realize of course that it's possible for someone to edit a blogger post. To cover my bases, I uploaded a screenshot of the post at 5:00 pm yesterday, and you can check the date the photo was uploaded.]

Part of the reason for the differences in the race for President was that I considered Anthony Goertz as a legitimate candidate. This was pretty much just a judgement call. My model typically lumps joke candidates and None of the Above into the same category which often works reasonably well. If I had put Goertz in that category, the race for President would have been predicted to be 43.9%-38.8% - only an average of 1.4% difference from the actual results.

To see how this compares with the results from before, take a look at this graph:

Comparing this to the previous iteration of this graph, it looks like the data points fit quite nicely. How cool is that?

Update!

At the request of some people, I've added this little excel web app. It will let you pitch any of the candidates of this election against each other in a fierce battle. You can have all eight competing if you really want!



Two points:
1) Please insert their name exactly as it was on the ballot. For instance, use "Josh Le" instead of "Joshua".
2) This used the budget values that they used for their actual election, so it may be a teensy bit unfair to pitch a candidate from an uncontested race against one that was contested.

Back to the original post:

Another thing I tried this year was to project the voter turnout before the election was finished. This graph shows the actual voter turnout and my projection of the actual voter turnout for each hour throughout the election:

In general there are two points in the graph where the projection significantly changes, and these take place between around 10:00-13:00 on each day. A cooler way of showing it is this:

Compared to previous years, this year had a much stronger showing on the second day relative to the first day, which is pretty cool. This could maybe be a sign of campaigning on voting days taking more of an effect.

The method itself of using the previous  trends to project voter turnout may not have been super accurate (my mid-day Wednesday projection was off by 1.4%), but it's possible that it may get better as it becomes more refined.

Anyway the end there got kinda rambly - sorry about that. Look at me still walking when there's science to do!

SU Model Results

This is a post that is going to have some numbers. Because of all the numbers, I'm going to have a disclaimer, and like all good disclaimers it will start in the form of a story.

Once upon a time, a washed-up old SU hack came up with an idea. He wanted to see if maybe there was a correlation between numerical inputs and voting results in elections.

So he made a machine. The machine ate numbers, and spat out numbers. 

And he saw all that he had made, and it was very good.

Well, maybe not. Using three years of election data to calibrate the formula is good for validating it. Having a model that takes old data and correctly predicts what happens in past elections is lovely, and the fact that it is as accurate as it is looks really good on paper. However, it is likely prone to over-fitting - small things that make the model accurate in older years may have been coincidences, or are not important in future years.

So with all that, let me say that I have no reason to believe these numbers are going to be particularly accurate. This is truly just me taking a formula that worked for previous elections, and using similar inputs to try to come up with numbers.

Note: One of the inputs is the budget summaries, available here.

President
Petros:  39.6% First-round votes
Saadiq: 35.1% First-round votes

Probability of Petros Kusmu winning: 70.2%

VP Student Life
Kevin:  35.5%
William: 47.9%

Probability of William Lau winning: 97.8%

Good luck...

A Message to Science Students

If you're a science student and you voted No on Sci5, you've made a huge mistake.

Hear me out for a second. This is completely irrelevant to any of the proposals that Sci5 put forward. I don't care why you voted no - in fact the merits of Sci5 are completely irrelevant to your error.

Bylaw 8200 of the Students' Union governs the process for levying a fee on students. Specifically, in order for a FAMF like Sci5 to be implemented, it needs to pass two criteria during voting:

• A majority vote in a general election agrees with the fee, and
• At least 15% of the faculty voted (section 18).

 Let's assume that the voter turnout for science this year is going to be similar to last year's total SU turnout (21.6%). This satisfies the second point listed above, and then as long as more than half (about 10.8% of science students) vote yes, the referendum passes.

However, if the students who were opposed to Sci5 just didn't vote in the first place, the referendum would have failed unless there was an overwhelming majority in favor. In fact, in a situation like this, voting the way you want could hurt your cause, which violates one of the most basic criteria of voting systems.

Normally this wouldn't be a terrifically large issue. For instance, last year's Science FAMF referendum got a turnout of 30.6% - at that point more than 15% of students could have voted yes, and every vote against would have been necessary to defeat it. It also wouldn't be much of an issue if you didn't know what the turnout of the election was going to be. The SU, though, regularly updates students on voter turnout, which is almost asking for the system to be gamed.

Basically, if I was an organizing an anti-Sci5 campaign, the best strategy to tell possible voters would be to not vote, and wait. Wait until the Science turnout hits over 15%. Maybe it won't, and you'll have won. If it ever does, then I'd get out as much of the vote as possible to try to tip the scales.

You'll note that I waited until after the Science vote hit 15%, just because I do actually abhor voting systems that allow rigging like this (not that I'd be particularly likely to influence anything, of course). I've proposed solutions to this back when I was a councillor, and I truly hope this gets fixed in the future.

Vote Turnout Projection [Update Thu 08:00]

Current turnout projection: 20.0%

Here are two new graphs, updated as of this morning. If you compare the first one with the previous set of graphs, you can see that it generally followed the predicted trend in shape, but was just slightly lower than anticipated.

Original Post:

After last year's election, I mentioned that when you look at the trends in voting behaviour between various elections, they're almost always the same. In fact, if you overlaid them (and corrected for length of election and total number of voters), they were very nearly identical.

What's really cool about this is that if you know the turnout at any given time, you can try to project the total voter turnout by the end of the election (assuming that the pattern holds this year).

So I did that.

Check it out. I have two graphs. The first one shows the total turnout so far (solid line) and the project turnout based on the curve from previous elections (dashed line).

And the second one compares turnout to projected turnout.

What's really cool about the second one is that the projected value hasn't deviated by more than a couple percent hour-to-hour for the last couple of hours. This suggests that the trend from the last couple of years has been followed pretty closely for a while now, although the beginning of the day was slower than anticipated.


Stay tuned for updates!