Spurious correlation or unidentified causal link?
To eat cheese or order the beans with no cheese? that is the question to ask for a good night's sleep.
Odd “spurious” correlations have been a funny theme site and it has been a widely shared theme since the increase in autism - “Correlation does not equal causation.” True, but it does imply it because it is often showing some connection. The phrase is to remind a scientist to check for other variables - maybe some other change affected both conditions equally, so they appear to trend on the same pattern, but it is because of the third thing.
In looking at a few of the examples, from a nutrition perspective, I saw a causal link rather than a silly example of science error.
A fact check of the statistics agrees that there appears to be a strongly significant correlation between “Per capita cheese consumption in the US” and “Number of deaths annually of people who died tangled in their bedsheets.” Then in readjusting by “detrending” the trend, the data was insignificant. Well, if you take the pattern out of a pattern, then you will have no pattern - true.
Spurious Correlations: Death by getting tangled in bedsheets and the consumption of cheese! (Aris Spanos) | Error Statistics Philosophy, errorstatistics.com.
Related quote: “There are three kinds of lies: lies, damned lies, and statistics.” - Mark Twain
Skipping over the question of the honesty in detrending trends for evaluating the statistical significance in a pattern - the immediate thought I had was: “Oh, cheese, indigestion, heart disease, tossing and turning, bad night’s sleep” caused lights out for enough people to show a correlated pattern over a ten-year time period. The statistical review also said ten years was not enough data - true, more data is always helpful.
Many examples on the spurious correlations site are fairly silly, however this one makes sense to me - Dietitian Fact Check => Plausible
I happen to have more data on cheese consumption in the US, but not on mortality rate of tangled bedsheets. My data indicates a need to reformat my data - it was a page on my very first website. I didn’t have time to write this post, let alone rewrite a table - so
~~~ just skim through this to see what I am discussing after the next: ~~~~
Calcium available to the US from our average food supply's kg/capita/day whole milk, exc butter, and cheese was calculated in column I for years ranging from 1961 through 2007. During that time frame there was a drop in total milk use and an increase in cheese.
*Total fat and protein is not on this table but there was a shift towards more fat and calcium and less protein with the switch to more cheese and less milk on average. the cheese is a more concentrated source of calcium and fat.
* Total milk intake dropped between 1970 and 1980.
A: Year B: Milk, exc butter, (total) kg/ capita/yr C: Whole milk, kg/capita /yr D: Cheese, kg/capita /yr E: Milk exc butter (total) kcal/capita /day F: Whole milk, est. calcium provided
G: Cheese kcal/capita /day H: Cheese, est. Ca++ prov. I: Total Est. Ca++ /day
Year Milk-Total Wh Milk Cheese Milk-total Est.Wh.milk Cheese Est cheese Total Est.
kg/cap/yr kg /cap/yr kg/c/yr kcal/c/day calcium kcal/cap/day calcium calcium/day
A B C D E F G H I
1961 265.81 190.63 5.66 382.97 634 mg 48.20 111 mg 745 mg
1965 254.48 185.94 6.15 383.26 619 mg 53.11 121 mg 740 mg
1970 243.25 176.96 7.30 371.86 589 mg 63.37 144 mg 733 mg
1975 231.91 143.21 8.43 354.18 476 mg 76.06 166 mg 642 mg
1980 239.13 137.37 9.87 359.05 457 mg 91.21 194 mg 651 mg
1985 257.14 140.65 12.01 393.48 468 mg 113.96 236 mg 704 mg
1990 256.34 129.45 12.52 381.67 431 mg 120.89 246 mg 677 mg
1996 252.32 117.18 13.58 375.42 390 mg 132.73 267 mg 657 mg
2000 254.00 119.03 14.86 383.72 396 mg 144.58 292 mg 688 mg
2001 256.89 135.46 15.00 380.99 451 mg 146.23 295 mg 746 mg
2002 259.62 115.01 15.26 385.01 383 mg 148.33 300 mg 683 mg
2003 257.38 117.25 15.66 386.14 390 mg 148.57 308 mg 698 mg
2004 253.51 129.64 15.07 374.42 431 mg 149.49 296 mg 727 mg
2005 254.00 115.05 15.04 378.09 383 mg 150.52 296 mg 679 mg
2006 249.73 113.73 14.86 372.62 378 mg 147.99 292 mg 670mg
2007 253.80 128.27 14.67 373.14 427 mg 146.60 289 mg 716 mg
A B C D E F G H I
[www.faostat.fao.org] - quantities of for columns B, C, D, E and G
Column F was calculated from column C x 2.2(convert to pounds) x 2 (convert to cups) / 365 days x 276 mg calcium / cup (used NBD # 01077) = calcium per day from kg whole milk available per capita/year.
Column H was calculated from column D x 2.2 (convert to pounds) x 16 (convert to ounces) /365 days x 204 mg calcium / ounce (used NBD # 01009) = calcium per day from kg cheese available per capita per year.
Column I is the sum of columns F and H and reflects a farily stable calcium intake although the source shifted from milk to cheese between the 1970 and 1990. A slight decline in total milk occurred during 1970 to 1980.
***Cream does not seem comparable to the amounts given in the 1909-2000 data. In this system, 0.01 kg/capita/yr is the number given for yr 2000, compared to the 5 to10 pounds per year in the 1909-2000 data. Calories from whole milk plus Cheese groups seem to equal the Milk, excluding Butter (Total) fairly closely, so I used those two groups to estimate calcium available.
[zagat.visibli.com/share/gYQnF2]*no longer works
~~~~
What the mess shows in column D is that the US per capita cheese consumption has increased from 5.66 kilograms for every citizen (just an average, babies, elderly, everyone), increased to 9.87 kg by 1980, 12.01 by 1985, 14.86 kg by the year 2000, and peaked at 15.07 kg in 2004, but remained fairly close in average between 2000-2007.
It also shows that more Americans also stopped drinking as much whole milk, but total milk consumption has remained about the same. So total calcium intake should be higher. I made a mistake in using the kg of Whole milk as the calcium estimate instead of using the Total milk for column G. There is ~ 204 mg of calcium per 100 grams of liquid milk. I used fluid milk, fortified, line 15 - Calcium in milk, per 100g - Diet and Fitness Today, so milk in 1960 may not have had extra calcium added. I am also not sure if the Total Milk, means, all milk used to make cheese, yogurt and liquid milk.
Re-estimating a total calcium intake per day from the Total Milk kg column B provides a new estimate for calcium mg per day.
2013 - Total Milk excluding Butter, kg/capita/yr, 254.69 kg = 2546.9 (100 grams) x 204 mg calcium, divided by 365 days = 1423 mg calcium daily average with no wastage (which does happen, so actual intake would be less). (https://www.fao.org/faostat/en/#data/FBSH)
Even if that is including all the cheese and yogurt, that is really a lot of calcium. The US nutrient guidelines may be too generous, especially with frequently low magnesium levels. If the Total Milk was in addition to the Total Cheese than Column H would be added.
1961 -1486 mg calcium from milk plus 111 mg from cheese = 1597 mg calcium per day.
2007 - 1418 mg calcium from milk plus 289 mg from cheese = 1707 mg calcium per day.
That trend is a cardiovascular risk. While it might seem like a good thing to be drinking less whole milk overall, eating more cheese is even higher in saturated fats.
In the last post a linked article suggested that a cheese or dairy rich diet is associated with worse sexual health - likely because it is linked to worse cardiovascular health. Calcium is inflammatory and a dairy rich diet is more acid producing during digestion - more oxidative stress inducing in other words. Additional antioxidants and magnesium would be needed to help detoxify a diet rich in dairy products. Milk protein can be difficult to digest, and high fat meals can also promote indigestion. Increased sweating and muscle cramps during sleep might be likely.
Spurious correlations may just need more study - asking why is this pattern here? What is involved in this pattern at a larger level?
Rather than just mathematically detrending the pattern - look no more pattern!
The US recommends 1300 mg calcium per day for adolescents and teens, and pregnant or lactating teens and 1000 mg for adults and pregnant and lactating adults - this US Calcium factsheet for health professionals seems wrong or changed regarding the prenatal recommendation.
“In this systematic review of 1 randomized clinical trial and 152 observational studies on dietary patterns and all-cause mortality, evidence demonstrated that dietary patterns characterized by increased consumption of vegetables, fruits, legumes, nuts, whole grains, unsaturated vegetable oils, fish, and lean meat or poultry (when meat was included) among adults and older adults were associated with decreased risk of all-cause mortality. These healthy patterns consisted of relatively low intake of red and processed meat, high-fat dairy, and refined carbohydrates or sweets.
Meaning This review found that a dietary pattern with nutrient-dense foods was associated with reduced risk of death from all causes.” (1)
The writing is on the wall - cheese may not be the best snack for our heart health and possibly for a good night’s sleep. Too much or too little sleep habitually may increase cardiovascular risks. The sweet spot may be 7-9 hours of sleep per night. (2) Teens need more sleep, similar to toddlers, about ten hours per night might be ideal for their brain and body development needs.
Correlation does not equal causation, but it indicates a strong possibility worth evaluating further.
Themes that are pushed by many media sources is an interesting pattern to watch. I had noticed a pattern in the last decade of more joking about Spurious Correlations and repetition of the phrase Correlation does not equal causation - as if it in itself is a fact. That is not true. The statement is simply a reminder that correlation alone cannot prove a causal link, but it certainly doesn’t disprove a causal link either. It strongly suggests there is a link - so double check whether something else is involved or data error occurred, or a too-small sample set might be not showing a good representation of the whole group.
I also noticed a pattern of more harassment of parents of vak injured children and parents of children with autism over the last couple decades. First mothers were blamed for not bonding with their child with autism and now it is simply being normalized as a nifty neurodiverse way to be new normal. Sadly about 70% of people with autism are unable to function well enough to live on their own. That is not a great new normal, and the rate continues to increase.
Better phrasing: Correlation suggests causation but does not prove it.
I got out of bed this morning, and yesterday, and the day before - did the bed make me get up?* Will the bed make me get up again tomorrow? Or am I choosing to get out of bed because I am no longer tired?
*High tech alarm clock bed simply tilts vertically at the designated hour, dumping you on the floor.
Thinking is fun, and free!
Addition - I forgot a helpful bit of math - how much cheese is that a day?
2007 - 14.67 kg/capita = 32.27 pounds = 516.32 ounces = 1.4 ounces of cheese per day per capita - all US citizens, cheese eaters and non-cheese eaters and cheese waste, all averaged together, based on huge units of commerce for the nation.
1965 - 6.15 kg/capita - 13.53 pounds = 216.48 ounces = 0.59 ounces of cheese per day.
So how much is too much? Individual metabolism varies in how well calories are digested and used, but a rough estimate would suggest a half ounce a day was moderate and a couple ounces or more a day is not moderate. Maybe, from a larger view, some people probably ate many ounces a day in 2007, while others eat no cheese - and the group average is 1.4 ounces.
One small pizza ~ six ounces of cheese may be used on a standard 12 inch which many people might eat as a single serving meal.
Pizza example: 1 small pizza is ~ 535 grams (eatthismuch.com) and 100 grams of cheese pizza provides ~ 294 mg of calcium based on the percentage of 2000 calorie diet Nutrient Label data. The recommendation is 1000 mg for adults. That small pizza would have ~ 1573 mg of calcium. That may be too much calcium for heart health.
The (pizzamaking.com) forum thought about 6 ounces, 170 grams, on a small 12 inch pizza was a standard amount (more probably is on an Extra Cheese). 221.3 mg calcium in 28 grams of mozzarella = 1344 mg calcium - still a lot of calcium for one meal or one whole day.
Making foods with more veggies adds fiber and phytonutrients and cheese can be used more sparingly. Choosing strong flavored parmesan, sharp cheddar, or a Swiss cheese can increase the impact of a small amount in a mixed dish or sandwich.
Moderation is the secret to good health - one or two pieces of a nutrient dense food and have a salad with some beans, nuts, or chopped meat for additional protein. Meal planning is about balancing the varied nutrient needs into a meal. More vegetables are needed than tomato sauce. Onions, mushrooms and green peppers on the pizza would be a fiber and nutrient addition. Cheese and dairy are too rich in calcium to be the primary or only protein food in a diet. Vegetarian diets need beans, nuts and seeds too, for additional trace minerals.
Disclaimer: This information is provided for educational purposes within the guidelines of fair use. While I am a Registered Dietitian this information is not intended to provide individual health guidance. Please see a health professional for individual health care purposes.
Reference List
English LK, Ard JD, Bailey RL, et al,. Evaluation of dietary patterns and all-cause mortality: A systematic review. JAMA Netw Open. 2021;4(8):e2122277. Available at: https://doi.org/10.1001/jamanetworkopen.2021.22277 https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2783625
Covassin N, Singh P. Sleep Duration and Cardiovascular Disease Risk: Epidemiologic and Experimental Evidence. Sleep Med Clin. 2016 Mar;11(1):81-9. doi: 10.1016/j.jsmc.2015.10.007. Epub 2016 Jan 9. PMID: 26972035; PMCID: PMC4791534. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4791534/
Spurious Correlations: Death by getting tangled in bedsheets and the consumption of cheese! (Aris Spanos) | Error Statistics Philosophy, errorstatistics.com.
https://www.eatthismuch.com/food/nutrition/small-cheese-pizza,176380/
Two additional columns just to save my math:
A-Date B-Total Milk was used to calculate mg calcium per capita using 204 mg calcium per 100 grams fortified liquid milk and divide by 365 days gives a replacement for column F - calcium in the Whole Milk per capita.
Total milk calcium - Col B, in mg/day:
1961 -1486, 1965-1422, 1970-1360, 1975-1296, 1980-1337, 1985-1437, 1990-1432, 1996-1410, 2000-1419, 2001-1435, 2002-1451, 2003-1438, 2004-1416, 2005-1419, 2006-1395, 2007-1418 mg calcium per day
2013 - 1423 mg calcium per day from the kg intake annually of Total Liquid Milk excluding Butter in the US per capita