Design

LSD is an acute water-soluble drug. This makes dosing easy: take a single dose of 250μg, dissolve it in some water (refrigerated), and then consume 1/10th for a dose of 25μg. Some online anecdotes microdose daily, while others microdose every 2 or 3 days.

I ordered 2 tabs ⁠from VitaCat on Silk Road⁠:

He shipped them in an envelope in an airtight sealed plastic sheet; I kept them refrigerated, underneath a box to block light, and kept sealed until I opened it and used the first one for ⁠a trip⁠. The remaining tab was kept in the refrigerator until a week later. (Given that people routinely stash tabs in books and other places, I doubt that the week did it much damage.) The next morning, I added 25ml of distilled water⁠^⁠11⁠ into an airtight mason jar, dropped the blotter in, shook vigorously, refrigerated overnight, and removed the blotter the next afternoon.

Self-blinding a liquid is as straightforward as ⁠self-blinding pills⁠: 2 opaque containers (one marked; both kept in refrigerator) into which a milliliter dropper puts a dose of LSD water and a dose of regular water.

This suggests the following design: a randomized dose on day 1, followed by days 2 and 3 off, then on day 4, drinking the second container; on day 7, examine the containers recording whether active/placebo and finally, starting over as day 1 of a new pair of 3-day blocks. The power analysis (see next section) indicates >19 active days are desirable, so 190+μg are needed; this can be diluted into an evenly divisible amount like 19ml of water and then 1ml extracted each day. 3 days seems to be the maximum that most microdosing enthusiasts consider a dose active for, so there shouldn’t be issues with tolerance or carryover. I ran this design by Dr. Fadiman, who did not object. (This design can be seen as a crossover study⁠: repeated measures⁠ of the same subject serving as their own control. No washout period to protect against carry-over, though, since I did not expect noticeable effects 4 days out from a microdose, and my later analysis suggests there was none.)

VitaCat’s Mayan tabs were advertised as 250μg, and I diluted the remaining tab into 20ml of water, so nominally each 1ml dose would have 12.5μg of LSD, which is at least twice the level (5μg) a number of people have claimed benefits from microdosing, so even if the tab was overstated by 100μg, the dose should still be enough to produce any effects.

I took one additional step: temporarily suspending my ⁠lithium self-experiment⁠. There are many mostly-negative opinions online about the interactions of LSD & lithium⁠, although most commenters seem to be talking about therapeutic doses of lithium - which are 10-15x larger than the 10mg of lithium orotate which I may have been taking if it wasn’t a placebo week. (It later turned out that it was indeed a placebo week so my precaution was unnecessary.) But there is no point in risking the described negative effects or the weakening of the trip, so 2 weeks before, the experiment was paused. Obviously I can’t prove that this is sufficient, but I think it was enough: I was taking 10mg orotate every other day (on average) and the anecdotes which caused me to place it on hold were from people taking psychiatric doses which are closer to 500mg, and they generally seem to think that the effect of lithium didn’t last for months (which is consistent with the fairly fast metabolism of lithium in the body, ~24 hour half-life); as well, I stopped it weeks before the trip, which went fine, which was itself a week before starting microdosing. So for the lithium to have been a problem, one would have to postulate that the levels of lithium, already 2 orders of magnitude below those claimed to cause problems, did not noticeably interfere with the trip, but somehow did interfere with results spread over 6 months afterwards.

The listed benefits are to mood, productivity, and creativity. We might also want to check other metrics to see if any LSD benefit came at a cost. So I will use my usual metrics plus a new creativity one:

• Zeo sleep data (primarily: latency, total sleep, number of awakenings, morning feel)

  I expect little or no effect on these 4 metrics, or the more obscure ones like sleep composition (light/REM/deep percentages). Two-tailed tests.

• Spaced repetition⁠ performance

  Average grade of cards reviewed on a given day; I expect no effect or a benefit (the simulated annealing analogy sounds like it might also work for memory). One-tailed.

• mood/productivity self-rating: 1-5

  3 is a normal day, 2 below-average, 4 a very good day, 5 fantastic etc. The major prediction of the microdose theory is that mood/productivity/creativity will increase, so the final analysis should exploit our prior and use a one-tailed test that the ratings will increase (since higher=better).

• creativity rating: 1-3

  Similar. This is not a metric I currently track, but would be a new one. One-tailed.

• active/placebo prediction

  A prediction recorded at the end of each block on PredictionBook.com⁠; this is not a dependent variable, but a check for whether there is a reliably noticeable subjective effect.

• Before/after Openness to experience⁠ questionnaires

  Openness is one of the Big Five⁠ personality traits, roughly linked with creativity, interest in novelty and variety; while Conscientiousness⁠ (think self-discipline & hard-work) tends to increase with age, Openness seems to decrease. Personality traits, like IQ, are notoriously hard to change & predictive of many things about a person. MacLeanet al2011⁠, a rare RCT⁠ of psilocybin⁠, found that Openness was still increased a few percentage points >1 year after dosing. (While not that large in magnitude, MacLeanet al2011 compares it with the reduction in Openness over 4 decades and increase from successful antidepressants or substance abuse treatment.) The effect seemed to be driven by those reporting a mystical experience; despite identical starting Openness, those reporting a non-mystical experience seemed to see theirs fall. The study has ⁠many weaknesses, but was still the best such study I knew of up to that point. A later RCT on LSD, ⁠Carhart-Harriset al2016⁠, found large effects 2 weeks after 75μg on both Openness and optimism (d = 0.16/0.56 respectively). Schmid & Liechti2017⁠, on the other hand, saw a smaller increase which was not maintained 12 months later.

  Since to maximize effectiveness, any full trip should be taken before micro-dosing, this suggest 3 samples: before-trip, after-trip, after-micro-dosing.

  My first long Big Five survey result (using ⁠Personality Project) in early 2012_13ya put my Openness at the ⁠87^th percentile; the ⁠second, 2 days after the trip, was identical (87th percentile). Since I had no mystic experience, this is consistent with MacLeanet al2011’s analysis. A ⁠third followup in March 2013, after the microdosing experiment, put me at the 93rd percentile (needless to say, the confidence intervals⁠ for all 3 overlap due to the percentile ranking being based on relatively few questions: ~6 for each personality factor).

We don’t need to worry about time-of-day effects. The LSD microdoses were at similar times each day (right after getting up). All the dependent variables were generally recorded/done at consistent times, except for the spaced repetition scores since I sometimes reviewed in the afternoon, but I’ve already investigated time-of-day effects⁠ in my spaced repetition performance and they are trivially small both in my dataset & in a multi-million-review dataset drawn from thousands of Mnemosyne⁠ users.

Limitations

There are 3 major potential problems with this experiment: the results may not generalize beyond me, the tabs may not have contained substantial levels of LSD, and the LSD may have degraded while being stored.

Power Calculation

The descriptions are that the effects are strong & noticeable, so a large effect size⁠ suggests a small experiment; but with 6 metrics, we will want to correct for multiple comparisons⁠ (which need will lower the required p-value in our one-tailed tests). We’ll assume a large effect size, p = 0.01, and paired⁠/within-subject⁠ experimental design⁠^⁠13⁠:

library(pwr)
pwr.t.test(d=0.8,type="paired",power=0.80,alternative="greater",sig.level=0.01)
 
     Paired t test power calculation
 
              n = 18.47818

19 pairs means 19 active and 19 placebo doses, 38 days total. This does not seem unduly onerous, but note that we’re assuming LSD has a strong effect since if we cut the effect size in half to d = 0.4 (a medium-small effect) we need 66 pairs or 132 days! (If the effects were oversold so d = 0.4 was the correct estimate, and we went with 19 pairs, we would not have an 80% chance of detecting the effect but rather a 24% chance. Oh well. “We do what we can, because, we must - for the people who are still alive.”)

How we treat individual days will matter: do we average the data for each block of 3 days into a single data point, or do we treat each dose as resulting in 3 data points - the effect on the first day, the lessened effect on the second day, and the weakest effect on the third day? (And then the next dose.) The latter seems to be a better strategy, and so with 250μg we get 25 doses of 10μg, each dose is 3 days so as much as 75 days for active and a similar amount for placebo. (75 pairs gives us a similar power calculation all the way down to d = 0.4.)

In retrospect (after observing the trends which didn’t reach significance), since I specified 7 metrics, it probably would have been better to be conservative and specify a significance-level of sig.level=(0.05/7) or 0.0071 rather than sig.level=0.05/5 or 0.0100.

VoI

For background on “value of information” calculations, see the ⁠Adderall calculation⁠.

With background, design, and power calculation out of the way, we can calculate costs & values.

1. Cost of regular LSD microdosing

   A check of listings on the Silk Road 1⁠ indicates that 300μg can be bought for ₿5–6 or $67^$45₂₀₁₃; 300μg translates to 30 doses, or 90 days if I follow the 3-day dose pattern some recommend; the yearly cost of LSD is then 45 /90⁄365 = $274^$183₂₀₁₃. Discounted at 5% annually, the net-present-value/lifetime-cost of switching to LSD would be 183 / ln 1.05 = $5,611^$3,751₂₀₁₃. We can reuse this figure as the potential benefit of LSD microdoses, since if the effect is positive and large enough to notice, then it seems worth roughly 50 cents a day.

   There are 3 major categories of additional costs: the medical risk one runs during the experiment (principally mental), the legal risk, and the reputational risk.

   1. Schizophrenia and other issues:

      Earlier, I looked into research bearing on the relation of LSD and Schizophrenia⁠, and concluded that the evidence was most consistent with LSD having small health risks (that is, the damaged or ill sought out all sorts of drugs including LSD, but LSD did not cause the damage or at best caused the illness to surface earlier) with some correlations of psychiatric benefit from use⁠; this was over the general LSD population including brain-frying large doses, so the risk for microdoses ought to be much smaller. Much later it occurred to me that the rarity of using LSD along with sheer small size of LSD actually reduces the risk of consumption compared to many other products, simply because dangerous doses of many contaminants or poisons won’t fit: blotters reportedly max out at ~500μg. Compare that with, say, piracetam⁠ where people take <3g of powder daily; if the piracetam was just 0.3% contaminants, that’s as much as 1mg of the stuff - but a blotter can’t even hold 1mg of any substance, much less an active dose of some substance plus accidental contamination. Overall, I feel comfortable rounding it to zero. To estimate the possible impact on schizophrenia⁠, I looked into estimates of hallucinogen consumption and found that estimates ranged from 7-12% of the entire American population (~22-38 million people) had consumed at some point, which over something like a 60-year lifespan (teenager to very old adult) implies something like 600,000 hallucinogen ‘virgins’ per year (a figure echoed in one of Fadiman’s articles).

   2. Prison:

      Searching, I found a ⁠1994 federal Sentencing Commission document⁠ listing 83 sentences involving LSD that year (while searching I also found some information indicating that 400μg is considered 1 dose, so the microdose experiment would involve <1 dose!); news articles suggest many (most?) of those would be for sellers and manufacturers, so a further analysis would reduce the risk accordingly. In any case, even if we multiply by 50 for each state, 600000 / (50 × 83) = 145 suggesting a rough estimate of 1 in 145 chance of being convicted. Leniency could be expected for a first-time non-violent non-trafficking offender, so we could set the cost of this outcome at perhaps just $149,574^$100k₂₀₁₃, or an expected loss/cost of $1,047^$700₂₀₁₃.

   3. Social consequences:

      Some reputational effects are clearly negative: a use of LSD would be a big issue in some roles like running for elected office or being CEO of a large company (unless one is a secular saint like Steve Jobs⁠), especially since I would obviously be unable to make an excuse like “youthful experimentation” or “a momentary lapse of judgment”, this writeup demonstrating that my use was thoroughly premeditated; however, this does not bother me inasmuch as I am unfitted by both temperament and talent for such positions. (Something similar could be said about security clearances.) In many more conservative regions, it would be a substantial negative, but I am already a poor fit for such regions and I don’t know whether an LSD use would be very negative at the margin, or whether any of my experiments or essays have already done the damage.

      On the positive side, however: in some groups or regions like California (which may yet lie in my future), a use of LSD may be a net positive as it indicates valued qualities like open-mindedness. The lifetime consumption rate implies that something like 12% of the population would not judge me harshly for any use. That I have been careful and responsible in my experimentation, and had multiple purposes in addition to mere recreation, will lessen the offense for many people (but possibly would increase my guilt in the eyes of others). There are many famous LSD users like Bill Gates⁠, Richard Feynman⁠, Oliver Sacks⁠, ⁠Francis Crick⁠, Kary Mullis⁠ etc, but they seem to have discussed their drug use only after achieving mainstream success; worse, the famous users all seem to have been associated with the 1960s or with computers, suggesting tolerance for them may be a passing phase and younger people not so lenient (but on the other hand, marijuana legalization⁠ efforts are actually succeeding before and during 2012_13ya, suggesting that America may be more tolerant in the future).

      All considered: I am not sure what the reputational cost would be, nor do I have good ways to estimate it, and given the invisibility of many reputational costs - how do you know when someone didn’t offer something to you or how people are bad-mouthing you? - I may never know the actual cost to me (compared to, say, being arrested for LSD possession).

2. How much will the experiment cost to run?

   The procedure each morning uses up no more than 5 minutes so (38 × 5) / 60 × 8 = $37^$25₂₀₁₃ dollars of time. The sleep, spaced repetition, and mood/productivity data I already collect, and the creativity self-rating is another few seconds, and so not worth calculating. I’m not sure if I have an appropriate dropper/syringe available; that and other supplies might tack on another $15^$10₂₀₁₃. So 45 + 25 + 10 = 80 dollars.

3. Priors:

   The accuracy of the information ranges 20-80% depending on the effect size; since it’s inline with my previous knowledge of LSD and based on more than a handful of anecdotes, I would be somewhat optimistic about there being an effect, so a 50% chance the effect exists.

4. Value of Information

   Power times prior times benefit minus cost of experimentation: 0.2 − 0.8 × 0.5 × 3751 − 80 − 700 = −405 to +720. If we cut out the legal risk, the range is +295-1420 and the experiment well worth doing; our estimate of the legal risk and the effect size determine whether we find it worth doing.

Fadiman Comments

After I completed the experiment and prepared the data, but before doing any analysis (besides the prediction accuracy to check the blinding), as a form of result-blind reviewing⁠, I emailed Fadiman the following questions & asking permission to quote him, and he replied:

1. You’ve seen my planning and methodology write up, and I’ve attached the current version of my full writeup with everything I’ve added or changed since (mostly more background, a log of dates, that sort of thing) if you want to reread it. Do you have any objections to how I did it?

Most important: I am deeply impressed and pleased with your totally original and cleverly designed experiment. That you are able to add a double-blind to a one person self-study is a delight in itself. I am incredibly happy to see how many different ways you went at this. In terms of the general write up, I happened to be an ardent fan of a mix of data, personal reflection and observation. Scientific writing all too often does everything it can to make the subject under scrutiny as dull as possible in the writing, as uninteresting as possible, and heaven forbid, to include any genuinely real and personal commentary. So I love the way you did it exactly. The changes all seem to me expansions and improvements.

2. Have you or anyone else as far as you know, done a long blind/randomized/both self-experiment on LSD microdosing?

Short answer. No. Long answer: no one. Longest answer - the number of people are doing self-experiments with LSD and other psychedelics and even MDMA⁠ is expanding. No one is even approaching what you’re doing.

3. If the results turn out to be non-statistically-significant or have very small effect sizes, why do you think they might have turned out that way? Which reasons would you have most confidence in as an explanation of null results?

I do have some guesses as to why non-statistical-significance might occur, and that is putting aside just the obvious statistical possibility of using small sample sizes. I will be actually very content to see null results since it will force me and others to dig more deeply for possible reasons. So far, you are the first who might produce such results and certainly since you’re the first double-blind, it adds to the value of what you’re doing. As you probably know there’s a big controversy in the regular medical literature about how many serious drug studies of multibillion-dollar drugs and up with no [statistically-]significant differences a few years out from their sales launch and almost always from university-based studies, not being paid by the pharmaceutical company.

In psychedelic studies, 100% success rates are much harder to describe to non-psychedelic audiences used to more conventional and much lower success rates about almost everything.

One possible reason, that only you can answer, and you may have done so in the body of the text, but if so I missed it. I’m not sure that you took a couple of non-blind micro-doses beforehand to be sure that you could notice a difference with every placebo and positive intention going with you to notice the difference to establish a personal baseline of awareness of the substance. If that was not done, then it’s not at all clear that the very simple measurements used on a daily basis were sensitive enough to pick up days on and days off. The more usual studies from other people that I’ve seen, (none of them even attempting a double-blind or any blind at all,) are looking for a set of internal changes that they have previously experienced. These may be as subtle as eating a healthier lunch, doing a few more reps at a gym, being able to focus on the task an hour longer and so forth. I have a couple of write-ups in my book and when you get to them, you can see that we’re dealing with fairly intricate internal measures which you also cover in detail now and then in your personal commentaries.

That’s my favorite possible reason. Another is that, while some people are very aware on the day of the micro-dose, others are clear that it is a two-day experience, and a few even longer. If even a tiny residue of the micro-dose is still active at the time of the next dose, it might make it harder to distinguish and on from an off day.

However, I’m also aware that also for some people, the dose is too low.

The report in my book that I find most charming, a woman who’s been micro-dosing for about 8 years - except during her pregnancy - uses 20 μg, which for me would be perceptual. So an alternative reason, even though the least interesting one, is that your dose was too low. A few people had felt it was better to drop down to 7 or 8μg to not get too high. Human variability.

I’m sure I can come up with more complicated reasons, but you asked what my thoughts were before we completed the analysis and there you are.

4. Do you expect the results to turn out positive? (By positive here, I mean statistically-significant after multiple correction and d > 0.1.) How much so?

I did expect the results to turn out positive because there are real effects that have been reported, not only in my little studies, but for the past few thousand years in indigenous societies who of course have experimented with micro-doses and every other conceivable form of use.

As to how [statistically-]significant the statistical breakout would be - I’ve not really thought much about it. This was your study and your level of statistical sophistication clearly exceeds my own. I tend to be somewhat suspicious of any study where the statistics are fairly exotic, since I know from watching my graduate students PhD dissertations, that when really simple and almost intuitively obvious statistics don’t show results, they start to shop around until they find something that gives them a positive result even if it’s a statistic that nobody else seems to know about. I watch statistical consultants assure students that they will find something that will justify hiring them. Is this pure science as God and Descartes would have it? Not in my book but hey, the real world is never as clean as the data derived from it.

5. Are you willing to be quoted on any of the above?

Of course, you are free to quote me in anything you write, and I’m free to complain that I was quoted out of context or anything else I choose to say to cover my tail. But in this case, you did the work - you asked me to review the work - and therefore my review becomes part of the total package.

I did this for much the same reason as I wrote down the exact metrics and analysis in advance: to specify what results are expected, how one will interpret them, and to eliminate the temptation to fudge or modify or spin or self-deceive about any of it.

Analysis

+++ mnemosyne/mnemosyne/example_scripts/export_stats.py 2013-03-10 21:54:04 +0000
@@ -12,7 +12,7 @@
 data_dir = None
 mnemosyne = Mnemosyne(data_dir)
 
-for n in range(-10, 0):
+for n in range(-275, -1):
     start_of_day = mnemosyne.database().start_of_day_n_days_ago(abs(n))

./bin/python ./mnemosyne/example_scripts/export_stats.py \
     | cut -d ' ' -f 2 | sed -e 's/None/NA/' > ~/mnemosyne.csv

lsd <- read.csv("lsd.csv")
lsd$Date <- as.Date(lsd$Date, format="%m/%d/%Y")
# Zeo's CSV export silently omits missing days; look for them to hand-edit empty rows in
# which(!((as.Date("2012-06-09"):as.Date("2012-09-15")) %in% lsd$Date))
mnemo <- read.table("mnemosyne.csv")
lsd$Mnemosyne <- NA
lsd$Mnemosyne <- mnemo$V1
mp <- read.csv("mp.csv")
lsd$MP <- NA
lsd$MP <- mp$MP
creativity <- read.table("creativity.csv")
lsd$Creativity <- NA
lsd$Creativity <- creativity$V1
write.csv(lsd, file="lsd.csv", row.names=FALSE)

let predictions = [(True,  0.45),(False, 0.55),(True,  0.60),(False, 0.40),(True,  0.40),(True,  0.60),
                   (False, 0.40),(True,  0.40),(False, 0.50),(True,  0.40),(False, 0.50),(False, 0.40),
                   (False, 0.60),(True,  0.40),(True,  0.65),(True,  0.50),(False, 0.40),(False, 0.60),
                   (True,  0.60),(True,  0.75)]
logScore ps = sum $ map (\(result,p) -> if result then log p else log (1-p)) ps
 
logScore predictions
→ -13.4685
 
log 0.50 * fromIntegral (length predictions)
→ -13.8629

Graphing Data

Below are the 8 main dependent variables, depicted over time & colored by whether it fell in a dose 3-day block or control/placebo block. In descending order of importance:

The 5 sleep variables:

lsd <- read.csv("https://gwern.net/doc/personal/2013-gwern-lsdmicrodose.csv")
# filter out baseline
lsd <- lsd[!is.na(lsd$LSD),]
 
l <- lm(cbind(MP, Creativity, Mnemosyne, Total.Z, Time.to.Z, Time.in.Wake, Awakenings, Morning.Feel)
            ~ LSD, data=lsd); summary(l)
# Response MP :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   3.0339     0.0898   33.79   <2e-16
# LSD          -0.1430     0.1293   -1.11     0.27
#
# Residual standard error: 0.69 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0108,    Adjusted R-squared:  0.00197
# F-statistic: 1.22 on 1 and 112 DF,  p-value: 0.271
#
#
# Response Creativity :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   1.4068     0.0734    19.2   <2e-16
# LSD          -0.1159     0.1056    -1.1     0.28
#
# Residual standard error: 0.564 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0106,    Adjusted R-squared:  0.00179
# F-statistic:  1.2 on 1 and 112 DF,  p-value: 0.275
#
#
# Response Mnemosyne :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)  3.82558    0.01625  235.40   <2e-16
# LSD         -0.00958    0.02340   -0.41     0.68
#
# Residual standard error: 0.125 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0015,    Adjusted R-squared:  -0.00742
# F-statistic: 0.168 on 1 and 112 DF,  p-value: 0.683
#
#
# Response Total.Z :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   519.80       6.36   81.67   <2e-16
# LSD            14.28       9.16    1.56     0.12
#
# Residual standard error: 48.9 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0212,    Adjusted R-squared:  0.0125
# F-statistic: 2.43 on 1 and 112 DF,  p-value: 0.122
#
#
# Response Time.to.Z :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    26.58       2.05   12.95   <2e-16
# LSD             2.04       2.95    0.69     0.49
#
# Residual standard error: 15.8 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.00425,   Adjusted R-squared:  -0.00464
# F-statistic: 0.478 on 1 and 112 DF,  p-value: 0.491
#
#
# Response Time.in.Wake :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    21.93       1.80   12.17   <2e-16
# LSD             2.78       2.59    1.07     0.29
#
# Residual standard error: 13.8 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0101,    Adjusted R-squared:  0.00128
# F-statistic: 1.15 on 1 and 112 DF,  p-value: 0.287
#
#
# Response Awakenings :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    7.237      0.382   18.96   <2e-16
# LSD            0.635      0.550    1.16     0.25
#
# Residual standard error: 2.93 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0118,    Adjusted R-squared:  0.00297
# F-statistic: 1.34 on 1 and 112 DF,  p-value: 0.25
#
#
# Response Morning.Feel :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    2.593      0.100   25.92   <2e-16
# LSD            0.370      0.144    2.57    0.011
#
# Residual standard error: 0.769 on 112 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0557,    Adjusted R-squared:  0.0473
# F-statistic: 6.61 on 1 and 112 DF,  p-value: 0.0114
summary(manova(l))
#            Df Pillai approx F num Df den Df Pr(>F)
# LSD         1  0.142     2.17      8    105  0.036
# Residuals 112
 
# MP is the most important metric; what is the effect size (Cohen's d) here?
(mean(lsd[lsd$LSD==1,]$MP) - mean(lsd[lsd$LSD==0,]$MP)) / sd(lsd$MP)
# [1] -0.1782
(mean(lsd[lsd$LSD==1,]$Creativity) - mean(lsd[lsd$LSD==0,]$Creativity)) / sd(lsd$Creativity)
# [1] -0.1921

p.adjust(c(0.27, 0.28, 0.68, 0.12, 0.49, 0.29, 0.25, 0.011), method="BH") < 0.05
# [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE

png(file="~/wiki/doc/nootropic/lsd/gwern-totalz.png", width = 780, height = 680)
qplot(Date, Total.Z, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-timetoz.png", width = 780, height = 680)
qplot(Date, Time.to.Z, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-timeinwake.png", width = 780, height = 680)
qplot(Date, Time.in.Wake, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-awakenings.png", width = 780, height = 680)
qplot(Date, Awakenings, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-morningfeel.png", width = 780, height = 680)
qplot(Date, Morning.Feel, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-mnemosyne.png", width = 780, height = 680)
qplot(Date, Mnemosyne, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-mp.png", width = 780, height = 680)
qplot(Date, MP, color=LSD, data=lsd)
dev.off()
png(file="~/wiki/doc/nootropic/lsd/gwern-creativity.png", width = 780, height = 680)
qplot(Date, Creativity, color=LSD, data=lsd)
dev.off()

Microdose Effect Length

One final detail to go with the previous analysis is to take a look at whether there were issues with 3-day blocks being a bad choice.

I classified by hand each day in the relevant period by how many days distant it was from the preceding LSD microdose (that is, the first day of an LSD block is 0, the second day is 1, the third day is 2, and so on up to 7, and past that I just round down to 7). The idea is that this lets us ask for a linear fit relating MP on days which are n distant from a LSD microdose and see if there’s any apparent trend - it may be that we failed to see a statistically-significant relationship because we lumped in all days together.

model1 <- lm(MP ~ DaysSince, data=lsd); summary(model1)
# Residuals:
#     Min      1Q  Median      3Q     Max
# -1.0760 -0.8587  0.0171  0.1413  1.1413
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   2.8587     0.0986    29.0   <2e-16
# DaysSince     0.0310     0.0206     1.5     0.13
#
# Residual standard error: 0.712 on 173 degrees of freedom
#   (99 observations deleted due to missingness)
# Multiple R-squared: 0.0129, Adjusted R-squared: 0.00721
# F-statistic: 2.26 on 1 and 173 DF,  p-value: 0.134
 
plot(jitter(lsd$MP, 0.5) ~ lsd$DaysSince, xlab="Increasing time since a microdose", ylab="Mood/productivity")
abline(model1)

More generally, we can re-run the multivariate regression with days-since as another predictor and see if it adds anything:

l1 <- lm(cbind(MP, Creativity, Mnemosyne, Total.Z, Time.to.Z, Time.in.Wake, Awakenings, Morning.Feel)
             ~ LSD + DaysSince, data=lsd); summary(l1)
# Response MP :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   3.2090     0.2980   10.77   <2e-16
# LSD          -0.2847     0.2639   -1.08     0.28
# DaysSince    -0.0323     0.0524   -0.62     0.54
#
# Residual standard error: 0.692 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0142,    Adjusted R-squared:  -0.00358
# F-statistic: 0.798 on 2 and 111 DF,  p-value: 0.453
#
#
# Response Creativity :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   1.2435     0.2434    5.11  1.4e-06
# LSD           0.0162     0.2155    0.08     0.94
# DaysSince     0.0301     0.0428    0.70     0.48
#
# Residual standard error: 0.565 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.015, Adjusted R-squared:  -0.00273
# F-statistic: 0.846 on 2 and 111 DF,  p-value: 0.432
#
#
# Response Mnemosyne :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)  3.80753    0.05400   70.51   <2e-16
# LSD          0.00502    0.04781    0.10     0.92
# DaysSince    0.00333    0.00949    0.35     0.73
#
# Residual standard error: 0.125 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0026,    Adjusted R-squared:  -0.0154
# F-statistic: 0.145 on 2 and 111 DF,  p-value: 0.865
#
#
# Response Total.Z :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   545.47      21.01   25.97   <2e-16
# LSD            -6.49      18.60   -0.35     0.73
# DaysSince      -4.73       3.69   -1.28     0.20
#
# Residual standard error: 48.7 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0355,    Adjusted R-squared:  0.0181
# F-statistic: 2.04 on 2 and 111 DF,  p-value: 0.135
#
#
# Response Time.to.Z :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    32.69       6.80    4.81  4.8e-06
# LSD            -2.91       6.02   -0.48     0.63
# DaysSince      -1.13       1.19   -0.94     0.35
#
# Residual standard error: 15.8 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0122,    Adjusted R-squared:  -0.00562
# F-statistic: 0.684 on 2 and 111 DF,  p-value: 0.506
#
#
# Response Time.in.Wake :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)   26.155      5.978    4.38  2.7e-05
# LSD           -0.639      5.292   -0.12     0.90
# DaysSince     -0.779      1.051   -0.74     0.46
#
# Residual standard error: 13.9 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.015, Adjusted R-squared:  -0.00275
# F-statistic: 0.845 on 2 and 111 DF,  p-value: 0.432
#
#
# Response Awakenings :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    9.374      1.251    7.49  1.8e-11
# LSD           -1.093      1.108   -0.99    0.326
# DaysSince     -0.394      0.220   -1.79    0.076
#
# Residual standard error: 2.9 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0396,    Adjusted R-squared:  0.0223
# F-statistic: 2.29 on 2 and 111 DF,  p-value: 0.106
#
#
# Response Morning.Feel :
#
# Coefficients:
#             Estimate Std. Error t value Pr(>|t|)
# (Intercept)    2.181      0.330    6.61  1.4e-09
# LSD            0.704      0.292    2.41    0.018
# DaysSince      0.076      0.058    1.31    0.193
#
# Residual standard error: 0.766 on 111 degrees of freedom
#   (13 observations deleted due to missingness)
# Multiple R-squared:  0.0701,    Adjusted R-squared:  0.0533
# F-statistic: 4.18 on 2 and 111 DF,  p-value: 0.0177
summary(manova(l1))
#            Df Pillai approx F num Df den Df Pr(>F)
# LSD         1  0.142     2.15      8    104  0.038
# DaysSince   1  0.076     1.07      8    104  0.390
# Residuals 111

None of the estimates are statistically-significant, but the MANOVA suggests that there’s modest evidence that the amount of time since a dose matters. To summarize the results for using DaysSince as well in a neat little table:

Variable	Effect	p-value	Coefficient’s sign is…
MP	-0.03	0.54	worse
Creativity	0.03	0.48	better
Mnemosyne	0.00	0.73
Total.Z	-4.73	0.20	worse
Time.to.Z	-1.13	0.35	better
Time.in.Wake	-0.78	0.46	better
Awakenings	-0.39	0.07	better
Morning.Feel	0.08	0.19	better

Do we get a better fit if we omit days “too far” from the microdose, reasoning that the effects would’ve stopped? Looking just at MP to keep things simple:

for (i in 7:2) { print(summary(lm(MP ~ DaysSince, data=lsd[!(lsd$DaysSince>=i),]))); }

days removed	correlation	p
	+0.03100	0.13
7th	+0.00418	0.90
6,7	-0.00514	0.89
5,6,7	-0.05220	0.32
4,5,6,7	-0.06680	0.35
3,4,5,6,7	-0.17400	0.11
2,3,4,5,6,7	+0.00000	1.00

Obviously the p-values are meaningless in this application; more importantly, the reversal of coefficient sign and changing coefficient size & statistical-significance don’t suggest anything in particular to me about dose period.