Last modified: 2013-05-01 (finished). Epistemic state: log.

but it’s too late, my friend\ too late\ but never mind

all my trials, lord\ will soon be over

– all my trials (excerpt)

As a simple safety precaution, and even though I don’t have a history of escalating addictions1, I’m set up a Beeminder goal to limit my cigarette use. I mean, the financial budget already keeps it at a manageable level, but that’s what they all say, isn’t it?

Nonetheless, I did a simple cost comparisons2 for a certain (fairly representative) Rich European Nation with its enlightened “heavily tax those who cost you less and who need the stuff to self-regulate” drug policy:

  • decent cigarettes: \~0.25 Euro / 1mg nicotine
  • roll your own cigarettes: \~0.10 Euro / 1mg nicotine (all other tobacco products are somewhere between those two)
  • lozenges: \~0.06 Euro / 1mg nicotine (that’s for 4mg lozenges, which are easy to split in half, but if you want to actually dose 1mg, you’ll probably have to pay double)
  • e-cigarette: well… ok, I’m confused by how much usable nicotine they really contain, and what realistic costs are (including hardware), but my best estimate is something like \~0.10 Euro / 1mg3 (especially as prices are expected to go up thanks to upcoming regulations), which surprises me as I thought they were a lot cheaper. Am I missing something?

I currently don’t have any reasonable way to circumvent those taxes (Eastern European border, I miss ya), so it seems like I’m not actually spending a whole lot of premium on cigarettes. Assuming a 4mg/day maximum (which is a pretty hig upper bound of my current use), I’d have to pay up to \~7 Eurons/month for the cognitive benefits through lozenges (which also ignores that I skip days irregularly because I forget about the stuff, which at least helps keep tolerance down).

If I add, say, 10 cigarettes/week (which is a similarly high long-term estimate of my current use) and use lozenges otherwise for the same 4mg/day, I’d pay \~9 Eurons/months, not a lot more. (And I love tobacco smoke and discolorations anyway, so this is a reasonable cost just for the pleasant experience, but I’m weird that way. Always loved having smoker friends for their smell, even though I only got interested in the stuff itself a few years ago.) So once my current supplies run out, that’s exactly what I intend to switch to.

But why stick with 4mg? Maybe that’s too much, maybe not enough? I need a better way to quantify what benefits I get from nicotine (in terms of cognitive enhancement, reduction of ugh, regulation of schizotypal symptoms). So I thought about existing options to get data from:

  • I can’t use sleep data because my sleep patterns are very irregular and my Zeo data is basically worthless. Back when I tracked it regularly for a few months, I virtually never dropped below a ZQ of 90 and saw many anomalies (catastrophic night with a ZQ > 120 etc.), and now Zeo’s dead anyway (pbut).

  • Comparing “time spent on useful projects” is tricky because there’s a lot of noise, but it might be workable if I do it for a sufficiently long period of time and I’m tracking this anyway.

  • Anki seems too noisy to evaluate. What would I even measure? Answer time? Correctness percentage is (on average) held constant by the scheduling algorithm and it’s tricky to control for difficulty.

None of those sound too promising, so I’ve looked into useful short tests I can run once every day. I want the total test suite to be short (<10min) so I don’t get bored, so I’ll stick with these 3 tests for 2 minutes each:

  1. n-back accuracy score

    I remember vaguely from gwern-sensei’s FAQ that it’s currently unknown if there’s a significant difference between single and dual n-back, and because I’m lazy I’ll start with single n-back. I’ll add sound when I get a round tuit.

  2. Answer time on simple arithmetic problems (“3+2=?”)

    This is Seth Roberts’ favorite. Many data points per day, plausible mechanism, useful skill. What’s not to like?

  3. Stroop task answer time

    There have been several studies that have shown that schizophrenics have significantly increased Stroop latencies, so it seems like a useful proxy test for schizotypy. The traditional positive/negative symptom tests (besides applying to most people in a philosophy department) are way too imprecise, so this will have to do.

I’ve hacked together a quick script that takes care of the assignments and tests, added a simple Beeminder goal to ensure daily measurements and I’ll figure out how to analyze the data once I have some. (It’ll take a bit for them to plateau anyway, so I can still work out problems in the protocol as they come up.)

There’s of course still the issue of randomization and blinding. E-cigarettes are relatively easy to blind, but everything else seems hard (use herbal / low-nicotine tobacco that tastes similar?) to impossible (find similar mint lozenges?). So I can’t properly control for placebo, but I can still learn some information about dosage. Next time I buy new lozenges, I’ll also buy weaker ones because they all look and taste the same and so I can blind strength. Until then and in addition, I can randomize scheduling - I simply randomly do the tests first and then take the nicotine or the other way around. That still provides a baseline to compare the intervention with. Lastly, I’ve also re-instituted my detailed drug log (which I only needed to avoid some nasty interactions, mostly with MAOIs, which I haven’t taken in some time) so I can spot potential patterns.

Empiricism!


So it seems I’m getting back into statistics and how to read scientific papers, and hey, I have some old notes but how much can I’ve really forgotten? Let’s review some simple stuff. What was a p-value again, Wiki-sama?

In statistical significance testing the p-value is the probability of obtaining a test statistic at least as extreme as the one that was actually observed, assuming that the null hypothesis is true. One often “rejects the null hypothesis” when the p-value is less than the predetermined significance level which is often 0.05 or 0.01, indicating that the observed result would be highly unlikely under the null hypothesis. Many common statistical tests, such as chi-squared tests or Student’s t-test, produce test statistics which can be interpreted using p-values.

Oh for fuck’s sake.

  1. I mean, look at that graph. If being very inconsistent in everything I do is a superpower, drug use is probably the only area where it pays off.

  2. Ignoring patches, which I found really hard to dose, and gum, which everyone I know who tried to use it had issues with. I also find it difficult to take absorption rates into account, but the “1mg” number is my best guess of effective nicotine, taking the numbers on cigarette boxes at face-value.

    It’s also funny how the nicotine content for each product does not generally affect it’s price much (if at all).

  3. Raw nicotine content in the liquid isn’t informative as such because it ignores the absorption rate (for which I couldn’t find a good source), other components besides nicotine, and so on. I ultimately went with the quantity of liquid former smokers typically end up using, which seems like a more useful proxy. However, even there good correlations are rarely present and estimates vary a lot. My (not particularly informed) best guess, based on that one study with self-medicating schizophrenics and various forums with smokers, is that one medium-to-high-strength cartridge equals \~6-10 cigarettes.

    I did find decent estimates for lozenges (and they fit my personal experience), so those costs are more reliable.

blog comments powered by Disqus
dlog » daily log » walpurgiskater