I drink more than I should. I’m confident about this. But, how do I know? To make the statement precise, I recorded the number of alcoholic beverages I consumed for every day in 2020 and in 2021. __Here is the analysis and target from last year__, and below are the results for 2021.

First, a recap.

On the basis of some research __I reviewed in 2019__, the recommended maximum drinks/day for men is 2 drinks/day, and the thresholds for “Heavy Drinking” are 4 drinks/day or 14 drinks/week for men. Note that "a drink" is not A Drink, in the sense that I haven't been precise about adhering to the definition of a Standard Drink Unit, or SDU, in either year of data-taking; for this reason, I've labeled the data as e.g. "drinks/day" rather than "SDU/day" as I did last year. More on this at the end.

I'll discuss last year's results as I go, to see where I improved (or not). I plan to mostly follow the same analysis procedures as last year, adding some comparison statistics as well. At the end of last year's post, I set the following targets for 2021:

Goal 1:Below the threshold for heavy drinking in every month (c.f. 1/12 in 2020).

Goal 2:At least 50% total days sober (c.f. 28% in 2020).

Goal 3:Fewer than 600 total drinks (c.f. 858 in 2020)

In retrospect, Goal 1 is ambiguous, as I didn't set a "threshold for heavy drinking" on a monthly timescale. I think what I meant was, below the threshold *for each week out of the month* (succeeded at this in May 2020, hence 1/12 months). The other goals are very clear.

Let's see how I did.

### Daily Intake

Here is the data over the course of the year.

As before, red region represents heavy drinking (14 dw ok, 15 dw over), "mean" means the mean value, and commonest' (note the ') is the modal value *excluding 0*, which is often the true mode. Note that the axes are chosen to match last year's analysis for ease of comparison. The abbreviations dd = drinks/day, dw = drinks/week.

**In total, I consumed 625 alcoholic beverages in 2021, compared with 858, a reduction of 233. My target was 600, so I missed it by 25 drinks.**

Let's compare directly to last year.

Eyeballing it, it seems like the peaks above the red line are lower, there are fewer of them, there are more days with drinks=0. The mean is 1.71 dd, below the target of 2 dd for healthy drinking, and commonest' = 2 as well. All good news!

The sample standard deviation (stdev) was stdev=1.80 dd, with a standard error of the mean (sem) of sem=stdev/sqrt(N)=0.094 dd given the sample size of N=365 days. Therefore my mean daily drinking habit is about 3 sem units below the acceptable amount. So regarding the question: **Do I drink more than the recommended 2 dd on average?** I can rule this out at the <0.001 level, or 99.9%. (Note that last year, I botched this very simple calculation and miscalculated my t-score by an order of magnitude. Dumb! But the conclusions are unchanged.)

On the question, **Did I drink less than last year?** I can similarly rule this out in a very significant way at the <0.001 level.

Still, I drink more than 4 dd ("heavy drinking") a non-negligible fraction of days. The frequency of days drinking N[dd] number of drinks is frequency[dd] = N[dd]/N, given below.

Modal score was 0, and commonest'=2 rather than 3 in 2020.

About 7.7% of the time, I drank heavily with ≥ 4 dd (c.f. 16% in 2020);

about 25% I drank too much but not heavily, between 3-4 dd (c.f. 30% in 2020);

and the other 67% of days I drank in the safe range of 0-2 dd (c.f. 54% in 2020).

**My total fraction of 0 dd days was 43%, with a goal of 50%, so I missed this goal by a margin of about 25 days. In other words, if I had drank 0 dd on 13 more days, I would have succeeded.**

One other interesting note: the fraction of days with 1 dd *dropped* from 2020 to 2021. This means, as I noted last year, that when I drink one drink, I tend to keep going.

Now we break it down month-by-month.

Best months were January and June, which were intentional no-drinking months. Worst month was July in terms of mean dd, or possibly July+August in terms of highest peak of 7 dd and also # of days above 4 dd (4 for each of the two months). July and August were when I was in the US, so possibly this is related to a lack of routine, stress of traveling, etc. (not confident in any particular explanation).

### Weekly Intake

Let's look at data by week.

Definite improvement in staying below the red line, but still room to do better. The mean dropped from 16.36 dw to 12.02 dw, which is just below the target of < 14 dw. You can see big troughs in January and June, which were intentional no-drinking months. Interestingly, the highest peaks (i.e. worst weeks) seemed to come right after no-drinking months, indicating a kind of rebound effect, or "__restrict-binge__" cycle.

The 2021 weekly sample stdev = 6.60 dw, implying sem = 0.92 dw since N=52 (about 30% larger than last year!). **Therefore my drinking in 2021 was below last year's mean of 16.36 dw by nearly 5 sem units (p<0.001), and below the threshold of 14 dw at about 2 sem units (p<0.05 very marginally).**

Frequency of drink number:

In 2021, about 35% of weeks, or 18 weeks out of 52, I drank heavily (c.f. 72% or 37 weeks in 2020). There were 10 weeks with < 7 dw (c.f. 3 in 2020), and 29 weeks with < 14 dw (c.f. 13 in 2020).

In individual months, results varied. (Note that I only use the first 28 days = 4 weeks, rather than taking fractional weeks or shifting start dates.)

I spend all of March, April, and December right at or above the line. The highest peak by far was in July, with nearly 30 dw (that's more than 4 dd on average!); looking at my calendar, it seems I was in St. Louis (with my in-laws) and in Cincinnati (with my grad school friends), so that kind of explains it. In general, I drank a lot while visiting the US (mid-July to mid-August).

__In December I made a resolution__ to drink only socially, not on an ordinary night at home, and I failed this miserably during the middle of the month. I blame this on a stressful job interview I gave just recently, but in any case, my resolutions seem to fail to overcome life circumstances in any significant way.

**If I interpret Goal 3 from last year in the way I suggested at the top ("keep every week of a given month below the heavy drinking threshold"), then I succeeded in 3 out of 12 months (c.f. 1 in 2020). The target was 12/12, which was maybe unrealistic, but in any case I missed it by a wide margin.**

I don't have a target for drinks/month (dm), but here's the data.

(mean = 52.08 dm, sem=19.91 dm)

January was statistically-significantly low, p<0.05, whereas June was low but not stat. sig, p<0.1.

### Conclusions and Discussion

Let's revisit __the conclusions of last year,__ amend them if needed, and then add more.

1. When I drink a little, I tend to keep drinking; this is evidenced by the dip in frequency around 1 dd and the peak at 3 dd. It may be preferable to focus on increasing the number of days with 0 dd than to moderate on days that I drink; in 2020 about 1/4 days I was totally sober.

This still seems true, though now the modal value (excluding 0 dd) was 2 dd rather than 3 dd, which is consistent with good drinking habits. Over the past year, I found myself feeling bad for having 1 or 2 drinks, because I targeted a lot of days with 0 dd, but actually 1-2 dd is fine and I should be happy with that. This year's resolution will reflect that.

2. I drink when I’m stressed; this is evidenced by the peaks in January and September, when I was making big life decisions or carrying them out. I should make an effort to reflect on how I’m feeling, and cope in ways that are healthy (more meditation, take walks, talk things out, etc.).

I still believe this too, though I would add that I also tend to drink when I am outside of a regular routine, e.g. when traveling. I didn't take my own advice to substitute drinking with meditation, walks, etc., which I'll also try to revisit this year.

3. I have power to change my habits when I focus on it; this is evidenced by the good month I had in May. In May my habits were largely healthy, with a mean of close to 1 dd and <10 dw. There were still three days with dd>4, but for the most part it’s a template for a healthier routine.

Maybe. January and June were both "don't drink" months, and they were indeed the months when I drank the least. However, June drinking was not stat. sig.-ly reduced, so "no-drinking" maybe means in practice "drink a lot less and feel bad about it." Furthermore, __December was supposed to be__ a "modest drinking" (only socially) month, and it was by no means a good month; rather, it was probably among the worst.

Worse than that, there were apparent 'rebounds' after my no-drinking months Jan and June, in February/March and July/August. These might be circumstantial, but also plausibly are direct result of restricting and feeling bad about it. To a good approximation, I can model the drinks/month figure above as a consistent ~55 dm, plus a (one month restriction)+(two months rebound) cycle. (Note that I don't see this pattern in 2020 data, in the aftermath of my May no-drinking month.)

\\

Overall, in light of last year's goals,

__Goal 1:__** **B̶e̶l̶o̶w̶ ̶t̶h̶e̶ ̶t̶h̶r̶e̶s̶h̶o̶l̶d̶ ̶f̶o̶r̶ ̶h̶e̶a̶v̶y̶ ̶d̶r̶i̶n̶k̶i̶n̶g̶ ̶i̶n̶ ̶e̶v̶e̶r̶y̶ ̶m̶o̶n̶t̶h̶** Below the threshold of heavy drinking for each week of the month, for 12/12 months (c.f. 1/12 in 2020).**

I did this for only 3/12 months of 2021.

__Goal 2:__** At least 50% total days sober (c.f. 28% in 2020).**

I did this for only 43% of days of 2021.

__Goal 3:__** Fewer than 600 total drinks (c.f. 858 in 2020)**

I drank 625 total drinks in 2021.

On one hand, I failed all three goals. On the other hand, the improvement was so great that I still count it as a successful year.

I drank __233 fewer drinks__ than last year!

I was sober for __55 more days__ than last year!

I drank heavily way less!

These are all good things that I want to build on for next year.

### The Problem of Imprecision

Now, for a bit of bad news. (If you don't care about precision of my drink calculations then you can skip to the next section, where I summarize the conclusions and outline goals.)

The truth is, I haven't been totally consistent with what I call "a drink". __Recall__ that there is actually a standardized unit for alcohol consumption, known as a Standard Drink Unit, or SDU. In the US, an SDU is defined as 14 grams of pure alcohol mass, so you can calculate the # of SDUs of a given drink by using the formula

where V is the volume of the drink, ABV is the alcohol by volume (measured as a percent), and D=0.789 g/mL is the mass density of alcohol. This is normalized such that a 12oz can of 5% ABV beer, or a 5 oz glass of a 12% ABV wine, or a 1.5 oz shot of an 80 proof = 40% ABV liquor, all give 1 SDU. (The type of alcohol doesn't actually matter; I just included them with the size of a roughly standard pour of each.)

On the other hand, I've been using the shorthand, colloquial idea of what a drink is, which reduces to something like 1 drink = 1 can or bottle of beer = 1 glass of wine = 1 shot of liquor. This doesn't take into consideration the ABV content of any of the drinks, which can be highly varied; beer tends to lie in the range 3-6% ABV, but some are 7% or higher, and __(unfortified) wines vary between__ 6-16% ABV! Liquor is even also variable, both within- and between-liquor types; this site quotes these as the key range for the most common liquors,

...but this is on the high side compared to various __other sites__, so take with a grain of salt (like tequila!). The point is, the calculation can easily shift by a factor of 2, or even more, in calculating SDUs precisely.

Furthermore, I haven't been careful to measure the quantity of drinks either (again, "1 glass" was taken to be "1 drink"). I looked through my usual kitchen supplies and tried to estimate how off I was likely to be, and in which direction.

First, I sometimes have beer. The two beer cans I have in my fridge at the moment are 500 mL = 16.9 oz Asahi wheat beer, and a 12 oz IPA; both happen to have 6% ABV. I would have counted the IPA as 1 drink, but actually it is 1.2 SDU. For the 'tall boy' 16.9 oz cans, what I've been doing is counting 1 can as 1 drink (which should be 1.7 SDU), and if I drink 2 cans I counted it as 3 drinks (which should be 3.4 SDU, or an average of 1.13 SDU for what I called "1 drink"). In all these example cases, I'm underestimating, but if I just take a naive average I find that for beer "1 drink" is about 1.34 SDU.

Second, I sometimes have wine. Our wine glasses are small compared to standard ones, and if I fill almost to the top, it comes out to 200 mL = 6.76 oz, compared to a standard wine pour of 5 oz; given that I don't usually fill the glass so high, maybe it's fine, but let's take it as a standard.

The wine I happen to have in my cupboard are a 6% ABV sangria, an 11.5% sparkling, a 13% Cabernet Sauvignon, and a 15.5% cheap sake that I treat like wine. One full wine glass = 200 mL of each of these would amount to 0.68, 1.30, 1.47, 1.75 SDU (respectively). Omitting sangria (which is very rare for me to drink), a naive average implies "1 drink" of wine is about 1.51 SDU.

Third, I sometimes enjoy liquor (whiskey is my favorite). I usually drink it with two ice cubes in my favorite green glass, which I made myself at a pottery shop; subtracting the ice cubes, it holds about 70 mL = 2.36 oz of liquid, compared with a standard shot size of 1.5 oz.

I like having a few different bottles with different flavors at any given time, and to share with friends if they like. Currently I have a 25% ABV shochu (Japanese liquor), a 40% scotch/tequila/Grappa (Italian liqueur), and a 44% Japanese whiskey; assuming my standard size of 2.36 oz in my favorite cup, this gives 0.98, 1.57, 1.77 SDU (respectively). A naive average gives 1.44 SDU.

If I assume I drink as much beer as wine as liquor, I can average the averages above, which gives (1.34+1.51+1.44)/3 SDU = 1.43 SDU. By this measure, my 625 "drinks" of this year was maybe more like 1.43*625 = 894 SDU!

After the 1.43 multiplier, here is the daily result:

...and for weeks:

...and for frequency:

With the precision of SDU restored, the data looks worrying again. So while I've made big gains this past year, there's still a lot of work to do. By this measure I only achieved Goal 1 in one month (January).

### Goals for 2022

As before, I'm using this opportunity to set some goals for 2022, which I'll assess again with similar data. Some key changes in my thinking:

I will be precise this year, recording not just "drinks" but the volume and ABV of each drink, so that at the end of the year I have a precise measure of SDUs. I hope this won't be too taxing; I'm developing a spreadsheet to make this as simple as possible (contact me if you want a blank copy for yourself). I'll compare 2022 to 2021 using the SDU unit, assuming the 1.43 multiplier on 2021 data.

This year, I wanted to achieve >50% days sober, so any day in which I drank any alcohol I felt kind of bad about it. But health-wise, any day in which I have "1 drink" should be fine, in the sense that 1-2 SDUs is not unhealthy by the measures I'm using. Because "1 beer" is often >1 SDU, and "1 whiskey" in my favorite glass is 1.5-1.8 SDUs, I'll round to <2 SDU as a marker for a daily target.

And with that, my goals for 2022:

__Goal 1:__** Below the threshold of heavy drinking (≤14 SDU/week) for each week of the month, for 6/12 months (c.f. 1/12 in 2020)**

__Goal 2:__** At least 60% of days (219 days) with <2 SDU/day (c.f. 50% in 2020)**

__Goal 3:__** Fewer than 600 total SDUs (c.f. 894 in 2020)**

As with last year, I'm committed to starting with a Sober January, with the hope that my knowledge of the rebound effect will help me avoid it. We'll see.

So here we go.

## Comments