Posts by JonathanM

Last ←Newer Page 1 2 3 4 5 Older→ First

  • Hard News: Gower Speaks,

    I actually don’t think you could do it that easily, because seats are distributed proportionally to parties over 5% or with electorate seats.

    Correct, but I think you could present "which combinations of parties could form government" reasonably enough. A Monte Carlo simulation of the seat distribution from the poll results would get the distribution of potential seats, and from there look at combinations in each case**, given that there's not really all that much doubt about who could go with who, other than with Winston. So a few scenarios and their chances could be presented, rather than the single "if an election was held today, it would look like this".

    Basically it would be: There's a K% chance that National/Act/UF/Maori will be in, an L% chance that Labour/Greens/Mana will be in, and a M% chance that NZFirst will be in parliament available for coalition talks.

    ** I suspect you don't need to care much about electorate votes as they only really effect the minor parties, and currently they're all showing little chance of coat-tailing.

    Since Jul 2012 • 64 posts Report

  • Hard News: Taking a very big gamble, in reply to Sofie Bribiesca,

    Yeah, I dunno whether the religious aspects of the SA would interfere with their ability to rehab folk from problem gambling or not. It might well be problematic for those problem gamblers that are particularly anti-religious to not be interested in rehab from the sallies, but I think we need to be careful about tarring them with the religion==bad brush unless there's evidence it's causing harm to them being able to deliver the rehab? Definitely something that should be looked into carefully.

    Since Jul 2012 • 64 posts Report

  • Hard News: Taking a very big gamble, in reply to Sofie Bribiesca,

    I presume you're aware, but to be clear, Seventh Day Adventist != Salvation Army.

    Since Jul 2012 • 64 posts Report

  • Hard News: What Hekia Parata actually said,

    The measuring of value added seems to me fraught with difficulty, particularly if you're wanting it standardised yet comprehensive (these being usually at odds).

    Still, if we assume this can be done, then it seems to me that estimating the value added per-teacher, given school effects, socio-economic effects and the like might be doable. There'd be quite a few assumptions involved. It would need to be adjusted for the multiple confounders that we know effect education progress that are unrelated to the teacher. We'd have to assume that the average teacher for each covariate pattern would achieve similarly, so we could then estimate the value-added by a particular teacher over and above that average. We'd need data from multiple years on each teacher on which we'd like to estimate 'value added performance'. We'd need a bunch of assumptions about how confounders interact in order to have a good model upon which the teacher-effects may be estimated.

    If we assume we can do all that, we'd end up paying them according to how well they do compared to other teachers in the exact same environment. i.e. it may well be that teacher A in environment B might get more money than the exact same teacher in environment C, just because they adapt to the confounding factors better than other teachers might.

    Even if all this could be done, is the best pay scheme to use to begin with?

    Since Jul 2012 • 64 posts Report

  • Busytown: School bully,

    Thanks for this Jolisa. I suspect most here will support your position. I see two problems with performance pay. The first is the obvious one: How do you evaluate performance? It's so highly dependent on the students in any particular class that it would be difficult to objectively assess.

    The second is that if you tie pay to performance then you're specifically introducing a disincentive to any material not on the test. Teachers love to teach material that's really interesting and related to the material assessed, but isn't directly assessed. And students soak it up - often this is where really great examples come from and real learning opportunities exist for the student. However, this material doesn't necessarily benefit the student's immediate grades - it takes time away from material that is directly assessed, meaning the trade-off is heavier under a constant-assessment system.

    In my experience, students are already (under NCEA) looking more and more at what they need to tick the boxes to pass, rather than actively interested in learning. Not all of them, but it seems as if there's an increasing proportion of university students focusing on "what will be in the test, what will be in the final, what do I need to know to pass" rather than "what is important that I should be learning". Maybe it's rose-coloured glasses, but it seems that with the cohorts since NCEA this is increasing?

    This is a tricky nut to crack. I don't think anyone doubts that there's lousy teachers (as you say - there's lousy students as well) but how you fix that is a really hard problem. In my opinion the best thing for a kid is to have a range of teachers - that way, yeah, they may get one that isn't a good fit for them, but they'll have others that are a better fit. This seems to me to be something that our schools already do a pretty good job with.

    Since Jul 2012 • 64 posts Report

  • Hard News: Friday Music: Going Large,

    Distributing audio for final listening purposes in 192/24 is ridiculous. Distributing for remixing/remastering/further processing, that may well make sense, particularly if the individual tracks (i.e. pre-mix) are distributed.

    Why? Science. http://people.xiph.org/~xiphmont/demo/neil-young.html

    If anything, 48kHz/24bit would be the sweet-spot at providing just a little more dynamic range (not that we don't have a heap already at 16bit) while still covering what we can actually hear.

    The claims about the player itself are somewhat ludicrous. A zero-feedback amplifier usually implies zero global feedback, but almost always has local feedback in multiple amplifier stages. Further, by stating that feedback can only correct errors that have already occurred is completely missing the point about why feedback circuits work so well.

    But hey, if it makes you think it sounds better, then that's a solid investment nonetheless - we frequently underestimate the value of the placebo.

    Sidenote: Does it annoy anyone else when kickstarter is co-opted as an advertising medium where the people behind the products already have plenty of funds to make it happen?

    Since Jul 2012 • 64 posts Report

  • Hard News: The Mayor's marginal enemies, in reply to Jim Cathcart,

    So, yes, among the PA readership, I may be on the outer in how I think of Len Brown. Nevertheless, even though my views are more Glenfield than Ponsonby, I am still rather surprised that anyone is surprised that LB is disliked. It seems more a case that if you assign to either a red or blue tribe, you have a natural support base, regardless of your efficacy or substance.

    This isn't the point of the post to which these comments are attached though, right? It's primarily about the Herald's cheerleading of this issue, which essentially consists of making shit up.

    Since Jul 2012 • 64 posts Report

  • Cracker: Stoned in Charge,

    Thanks Damian - will take a look around the ICADTS site and see what I stumble across. Much appreciated.

    Since Jul 2012 • 64 posts Report

  • Cracker: Stoned in Charge,

    Interesting read Damian - thanks!

    Any chance you have a link or citation for the Canadian study on fatal crashes? An odds ratio of 40 sounds quite close to 5*8 (odds for cannabis * odds for alcohol), suggesting that either no interaction was included in the model, or that the interaction was small or non-significant, which would be an interesting result.

    As an example of why it's interesting, let's suppose that being under the age of 30 increases your chance of a fatality by 4. Then, though we'd say "Smoking cannabis increases your chance of being in a fatality by 5 times, but if you're under 30 it's 20 times", that betrays what is really going on - the effect of cannabis is 5 times regardless of age.

    Thus, in this case, it may well be 5 times regardless of whether or not you've also been drinking. The combined activities simply increase your risk because they're combined: there's no additional risk on top of the risk you get from both activities separately - no interaction of effects. That's surprising.

    Since Jul 2012 • 64 posts Report

  • Southerly: A Blog on Behalf of an…,

    One concern might be that jurors would take their lead from support staff, who would likely have longer positions and be more skilled on average than the jurors at interpreting legislation. The turnover rate of jurors might be enough for this to be minimised though?

    One might get an estimate of how effective it would be by repeated random sampling from the proposed age groups about "controversial" legislation before the final reading. The GSCB, Gay Marriage and Electoral Finance Act bills would have been interesting ones to see. My guess is that the first two would pass. Not sure about the latter.

    Since Jul 2012 • 64 posts Report

Last ←Newer Page 1 3 4 5 6 7 Older→ First