I Does this scatter chart really show a cause and effect?

[arydberg]

I tried to show a link between chicken consumption and obesity. What I came up with is on a web page. You can see it here:
http://www.maverickexperiments.com/chicken/chickenFat.html

My question is does the scatter chart at the end of this page valid?

 

[berkeman]

I tried to show a link between chicken consumption and obesity. What I came up with is on a web page. You can see it here:
http://www.maverickexperiments.com/chicken/chickenFat.html

My question is does the scatter chart at the end of this page valid?

Looks pretty random to me, as I would expect. You are not accounting for other variables in this -- you would need to somehow remove variations from other factors in order to implicate chicken consumption on its own, IMO.

 

[Windadct]

This only shows correlation - not cause and effect. For example - consider being obese and this causes the person to crave chicken.

Or if you looked at body builders - who in general eat a lot of skinless chicken breast, you would be saying chicken breast makes you buff.

 

[mfb]

Industrialized countries tend to have more obese people, and they tend to have a higher meat consumption. This is not chicken-specific. You probably get a better correlation if you plot total meat consumption.

But, as commented already, correlation and causation are different things. You get an even higher correlation if you plot CO2 emissions per person versus the fraction of obese men.

You can also get great correlations with things that are clearly without causal relation. This website has many of them.
The number of people who drown in a pool correlates with the number of films Nicolas Cage appears in. What causes what? ;)

 

[berkeman]

This website has many of them.

LOL

 

[Mark44]

I tried to show a link between chicken consumption and obesity. What I came up with is on a web page. You can see it here:
http://www.maverickexperiments.com/chicken/chickenFat.html

My question is does the scatter chart at the end of this page valid?

A couple of points in the scatter chart are interesting. The U.S. and Australia consume almost the same amount of chicken, per capita, but the U.S. percentage of obese people is almost 12 percentage points higher. Also, Poland and Myanmar (Burma, really) consume nearly identical amounts, yet Poland's share of obese people is 28.6% while Myanmar's share is only 3.7%. Clearly, there is something else going on besides the amount of chicken being eaten.

 

[OmCheeto]

A couple of points in the scatter chart are interesting. The U.S. and Australia consume almost the same amount of chicken, per capita, but the U.S. percentage of obese people is almost 12 percentage points higher. Also, Poland and Myanmar (Burma, really) consume nearly identical amounts, yet Poland's share of obese people is 28.6% while Myanmar's share is only 3.7%. Clearly, there is something else going on besides the amount of chicken being eaten.

Yes, something is going on here. But what it is, I'm not sure.
Anyways, in an attempt to comprehend what my t-shirt means, by this coming weekend, I decided to do a little data mining, and plotting...

variable____    R^2__     equation_(ax + b)
calories____    0.7178    0.022x - 46
rice________    0.6340    -0.15x + 29
all meats___    0.5959    0.25 + 6.1
emp-Agricult    0.5415    -0.54 + 30
cows________    0.4778    0.22x + 13
emp-Service_    0.4567    0.50x - 9.3
chickens____    0.4517    0.63x + 6.2
income______    0.2934    0.0005x + 11
ave temp____    0.2004    -0.53x + 29
fish________    0.0744    -0.21x + 26
leisure time    0.0504    -2.5x + 64
emp-Industry    0.0267    0.29x + 16
working time    0.0097    0.0041x + 21
vegetables__    0.0031    0.0077x + 21

My conclusions were:
1. Given that I have never studied statistics, these numbers are kind of gibberish, but somewhat entertaining.
2. A higher R² seems to indicate that the data is less random.
3. No matter how many vegetables you eat, it isn't going to make you skinny, if the other 13 variables are not in your favor.
4. I will never understand my t-shirt*.

------------------------------------
*Om's "Radical Bayesian" t-shirt:
The OP's hypothesis was that eating chickens makes you fat.
My current hypothesis is that nations with low rates of agricultural employment, and high rates of service employment will make you fat.
How on Earth do you put that into an equation? Do I just plug in the R squares?

ps. Sorry for the thread hi-jack, but I'm serious.

 

[arydberg]

A couple of points in the scatter chart are interesting. The U.S. and Australia consume almost the same amount of chicken, per capita, but the U.S. percentage of obese people is almost 12 percentage points higher. Also, Poland and Myanmar (Burma, really) consume nearly identical amounts, yet Poland's share of obese people is 28.6% while Myanmar's share is only 3.7%. Clearly, there is something else going on besides the amount of chicken being eaten.

There is no question that many other things are going on. My question is. Is there sufficient evidence to show that the chemicals and antibiotics used in the production of chicken may be partly responsible for obesity? Should this possible cause be suspected and studied.

 

[mfb]

The plot does not help answering that question at all.
Even if you could establish a clear link between chicken consumption and obesity (you cannot with the given data), it would not imply that antibiotics and "chemicals" (whatever that means - water is a chemical substance as well) are responsible, instead of the chicken itself.

 

[arydberg]

Industrialized countries tend to have more obese people, and they tend to have a higher meat consumption. This is not chicken-specific. You probably get a better correlation if you plot total meat consumption.

But, as commented already, correlation and causation are different things. You get an even higher correlation if you plot CO2 emissions per person versus the fraction of obese men.

You can also get great correlations with things that are clearly without causal relation. This website has many of them.
The number of people who drown in a pool correlates with the number of films Nicolas Cage appears in. What causes what? ;)

What causes that you ask. Simple. The number of films Nicolas Cage appears in is cherry picked from thousands and perhaps millions of sets of data. It is chosen as it happens to match the data for the number of people who drown in a pool.

To me the statement "correlation does not imply causation" is akin to Newton "finding a beach pebble brighter than most" and then throwing it away. Newtons genius lies not in finding the pebble but in seeing in it what could be derived from it.

 

[mfb]

The number of films Nicolas Cage appears in is cherry picked from thousands and perhaps millions of sets of data. It is chosen as it happens to match the data for the number of people who drown in a pool.

That is the point.
Why do you think chicken is not cherry-picked from all the foods you could look at? Especially as you would expect a correlation between most foods and obesity, simply due to regional differences.

 

[arydberg]

Chicken is the only food I know of where we have managed to double it's growth rate. To make my case clearer I've added more graphs. Look again.
http://www.maverickexperiments.com/chicken/chickenFat.html

Al

 

[OmCheeto]

Chicken is the only food I know of where we have managed to double it's growth rate. To make my case clearer I've added more graphs. Look again.
http://www.maverickexperiments.com/chicken/chickenFat.html

Al

Chickens...
hmmm...
Two years ago, I built a garden, and had more tomatoes than I knew what to do with.
Last year, I got one tomato.

My growth rate went down by about a factor of 100.

But previous to that, I tried to grow tomatoes in my "alt-garden".
The growth rate there was consistent with my last years growth rate in my "neo-garden".

Conclusion?

Water and feed(solar and nutrient-wise) your tomatoes, and you will get lots.

As to your chicken double growth rate assertion?

Ummm...

Do you have any sources for that?

 

[arydberg]

Chickens...
hmmm...
Two years ago, I built a garden, and had more tomatoes than I knew what to do with.
Last year, I got one tomato.

My growth rate went down by about a factor of 100.

But previous to that, I tried to grow tomatoes in my "alt-garden".
The growth rate there was consistent with my last years growth rate in my "neo-garden".

Conclusion?

Water and feed(solar and nutrient-wise) your tomatoes, and you will get lots.

As to your chicken double growth rate assertion?

Ummm...

Do you have any sources for that?

• • Sure here:
• Conventionally bred broiler chickens: Most of the chicken available in the stores today comes from flocks that grow to market weight in about 48 days on average, using fewer natural resources – therefore more sustainably. Compared to 25 years ago, today’s chickens now require seven percent less feed per pound to grow. Considering our national broiler flock eats about 57 million tons of feed per year – that is a lot of resources saved.

• Slow growth/Heritage broiler chickens: “Slower-growing” chickens or “Heritage breeds” are chickens that can take almost twice as long to reach market weight – about 81 days typically – because they do not convert feed to muscle as quickly. Because of this, these breeds require more feed, fuel, water and land per pound of meat to sustain their growth. As such, these products are typically 3x more expensive than their counterparts.

from

http://www.chickencheck.in/faq/difference-faster-slower-growing-chicken/

Al

 

[mfb]

Chicken is the only food I know of where we have managed to double it's growth rate.

We managed to more than double the yield of basically every crop. So what?

To make my case clearer I've added more graphs. Look again.
http://www.maverickexperiments.com/chicken/chickenFat.html

Even 100% correlation would not make a case at all.

Unrelated: Calculate the correlation instead of putting your plot next to a random 50% correlation plot.

 

[Merlin3189]

Looks pretty random to me, as I would expect.

Interesting difference in perspective. To me it looked a pretty good correlation, well worth doing the stats* (which would have been without the aid of a mechanical calculator.) But that was the perspective of a psychology student, rather than a P/S/M/E graduate.

* to evaluate the correlation, NOT to show any causal link.

 

[arydberg]

We managed to more than double the yield of basically every crop. So what?
Even 100% correlation would not make a case at all.

Unrelated: Calculate the correlation instead of putting your plot next to a random 50% correlation plot.

You are mixing up growth rate and yeld.

 

[arydberg]

Interesting difference in perspective. To me it looked a pretty good correlation, well worth doing the stats* (which would have been without the aid of a mechanical calculator.) But that was the perspective of a psychology student, rather than a P/S/M/E graduate.

* to evaluate the correlation, NOT to show any causal link.

Thanks for the response. It appears to me there is some correlation and that a further look is justified. Thats all.

 

[mfb]

You are mixing up growth rate and yeld.

I am aware of the difference. So what? You propose a causal effect that is not backed by any experimental data or any plausible biological mechanism shown so far. Why do you expect an effect with chicken and not an effect from massive increases in crop yield?

 

[Aufbauwerk 2045]

You ask an important question, and your hypothesis may be correct.

However, as others have pointed out, you haven't quite proved that eating food with more growth hormones causes more obesity. You are not plotting obesity vs the consumption of growth hormones, only vs. the consumption of chicken. So we don't really know how much, if any, growth hormones contributes to obesity, just based on your presentation. For one thing, the amount of growth hormones present in chicken may vary from country to country.

Perhaps you could get data that would allow you to make two graphs. One would show obesity vs. consumption of organically grown chicken, and the other vs. consumption of chicken who are raised using growth hormones.

I think weight gain or loss in general is well understood. There is actually an equation that predicts weight loss based on your age, gender, current weight, amount of physical activity, and projected calorie intake. The version of this equation I used does not distinguish between calories from fats, carbohydrates, or protein. It also does not distinguish between foods with or without growth hormones.

I went on a very simple calorie restricted diet where I ate whatever I felt like eating, but limited my daily calories. I actually drew a straight line when I began the diet, based on the equation. I compared my daily weight with the prediction line. The experimental result matched the theoretical result very well indeed. Sorry, I don't have the link to the equation, but it should be easy to find.

In any case, I like your question because I am one of those who prefers organic food and stays away from the growth hormones as much as possible.

I think statistics is not an easy subject. I found it all a bit confusing until I read Perry Hinton, Statistics Explained.

 

[OmCheeto]

You ask an important question, and your hypothesis may be correct.

However, as others have pointed out, you haven't quite proved that eating food with more growth hormones...

You are the first person to mention "growth hormones".

per wiki:

Poultry Farming: Growth hormones

Hormone use in poultry production is illegal in the United States. ... Several scientific studies have documented the fact that chickens grow rapidly because they are bred to do so...​

also per wiki:

Corn/Maize: History

Before they were domesticated, maize plants only grew small, 25 millimetres (1 in) long corn cobs, and only one per plant. Many centuries of artificial selection by the indigenous people of the Americas resulted in the development of maize plants capable of growing several cobs per plant that were usually several centimetres/inches long each.​

Corn appears to be somewhat longer than 25 millimetres nowadays. hmmmm... Does eating modern fat corn make us fat?

Also, I spent some time studying the WHO obesity data, and it looks to me like obesity is 90% genetic.
The 9 most obese nations in the world are all in the same general location

Filter: WHO region: Western Pacific

48.6 Cook Islands
46.6 Niue
46.0 Nauru
45.4 Palau
40.5 Tonga
39.2 Samoa
38.0 Marshall Islands
37.4 Tuvalu
35.4 Micronesia
​

I didn't bother seeing how much chicken they ate.

According to CNN:

The epidemic began through the tropical region turning its back on traditional diets of fresh fish and vegetables and replacing them with highly processed and energy-dense food such as white rice, flour, canned foods, processed meats and soft drinks imported from other countries.​

 

[Aufbauwerk 2045]

You are the first person to mention "growth hormones".

You are correct. As @arydberg says, "My question is. Is there sufficient evidence to show that the chemicals and antibiotics used in the production of chicken may be partly responsible for obesity?"

If you go to http://www.maverickexperiments.com/chicken/chickenFat.html you find that the hypothesis is that there are growth-promoting chemicals in chicken that cause the people who eat chicken to become obese. To quote, "We have developed chemicals and antibiotics that cause the growth rate of chickens to almost double." But you are right because this does not mention growth hormones. My mistake.

 

[SlowThinker]

This thread doesn't make much sense. Try plotting fatness vs. number of cars per person, or just vs. GDP per capita. I bet you'll get a much better correlation.

AFAIK chicken grow fast because they have ideal conditions. I know for a fact that temperature is regulated to $0.1^oC$ within an ideal temperature profile, probably also lighting does not resemble natural daylight.
Anyway, meat is meat, there are no homeopathic imprints or toxic substances in it.

It is probably true that developed countries eat too much of *any* meat, though. But I doubt statistics can shed some light on this, because there are too many other variables. You need a controlled experiment where the only difference is the amount of chicken eaten, over at least a decade, in a group of at least several tens of people.

 

[arydberg]

I have said what I have to say. Four people disagree totally. Three seem to be in partial agrement. Of course science is not done by vote but thanks for all the replies.

 

[NTL2009]

Partial agreement with what?

 

[arydberg]

Partial agreement with what?

A link between chicken consumption and obesity.

 

[NTL2009]

I have said what I have to say. Four people disagree totally. Three seem to be in partial agrement. Of course science is not done by vote but thanks for all the replies.

Partial agreement with what?

A link between chicken consumption and obesity.

One of us (or maybe both, to include both possibilities) is really bad with numbers. Where do you see three people in partial agreement?

One poster, (Merlin3189) said he saw some correlation, but also repeated that correlation does not imply causation, and your further mention of growth "chemicals" hasn't been accepted by anyone, other than one poster incorrectly mentioning the use of growth hormones (which are not used, illegal and also not economical, which means there is no incentive to cheat).

Why are you trying to assign cause to something so tenuous?

 

[Mark44]

One poster, (Merlin3189) said he saw some correlation, but also repeated that correlation does not imply causation

I agree completely. Since you (@arydberg) titled this thread "Does this scatter chart really show a cause and effect?", it seems that you are confusing what is merely a correlation between obesity and chicken consumption as a causative factor.

With the data shown, one could also conclude, with the same amount of justification, that obesity causes people to eat chicken.

 

[arydberg]

One of us (or maybe both, to include both possibilities) is really bad with numbers. Where do you see three people in partial agreement?

One poster, (Merlin3189) said he saw some correlation, but also repeated that correlation does not imply causation, and your further mention of growth "chemicals" hasn't been accepted by anyone, other than one poster incorrectly mentioning the use of growth hormones (which are not used, illegal and also not economical, which means there is no incentive to cheat).

Why are you trying to assign cause to something so tenuous?

I asked a question. You gave me your answer. Why do you keep coming back at me? Am i not allowed to ask this question because of some crazy saying like.

"correlation does not prove causality"

If you wish to believe the saying above it is your business. Good luck to you. As for me i will continue to question and doubt everything I see and hear. That is who I am and it has served me well.

 

[arydberg]

I agree completely. Since you (@arydberg) titled this thread "Does this scatter chart really show a cause and effect?", it seems that you are confusing what is merely a correlation between obesity and chicken consumption as a causative factor.

With the data shown, one could also conclude, with the same amount of justification, that obesity causes people to eat chicken.

I never said it was a causative factor.

 

[NTL2009]

I asked a question. You gave me your answer. Why do you keep coming back at me? ...

Because your answer didn't make sense to me. Where are the three posters who "partially agreed" with you?

You posted to a forum. Questions are asked, answered, and discussed. If you don't like that, I'll give the same advice I give others in this situation: start your own blog, where you can control the comments, and allow only those that agree with you.

Am i not allowed to ask this question because of some crazy saying like.

"correlation does not prove causality"

If you think that saying is "crazy", there is little hope you will make progress with investigations into data.

If you wish to believe the saying above it is your business. Good luck to you. As for me i will continue to question and doubt everything I see and hear. That is who I am and it has served me well.

Question and doubt are good tools. But at some point you need to accept something, or you can't get off the starting line. And it is pretty easy to accept that "correlation does not prove causality", it's been discussed seriously and specifically in this thread, and with some more light-hearted (but still with a serious point) illustration with that website of various correlations that are clearly just coincidence.

To re-hash, the correlation in that data is not strong, maybe not even weak. You should gather much more data before even attempting to take anything from it. Some has been suggested (meat consumption in general, calorie consumption, activity levels, age, gender, etc). If you think that single graph is curious enough to raise the question, fine. But that means go on to these other areas, and see if anything holds. One chart is barely a starting point.

 

[Mark44]

I never said it was a causative factor.

No, but as I said in my previous post, which you quoted, you seem to be thinking that eating chicken causes obesity.
Thread title "Does this scatter chart really show a cause and effect?"
As has been stated numerous times in reply, the answer is "No."

 

[arydberg]

No, but as I said in my previous post, which you quoted, you seem to be thinking that eating chicken causes obesity.
Thread title "Does this scatter chart really show a cause and effect?"
As has been stated numerous times in reply, the answer is "No."

The answer is not "NO" The answer is that no one really knows. To you the answer is No. I can accept that for you as well as others. What I cannot accept is your disagreement with me because i do not think like you would like me to think. I have no planes to change . Could we please just agree to disagree.

 

[arydberg]

Because your answer didn't make sense to me. Where are the three posters who "partially agreed" with you?

You posted to a forum. Questions are asked, answered, and discussed. If you don't like that, I'll give the same advice I give others in this situation: start your own blog, where you can control the comments, and allow only those that agree with you.

If you think that saying is "crazy", there is little hope you will make progress with investigations into data.
Question and doubt are good tools. But at some point you need to accept something, or you can't get off the starting line. And it is pretty easy to accept that "correlation does not prove causality", it's been discussed seriously and specifically in this thread, and with some more light-hearted (but still with a serious point) illustration with that website of various correlations that are clearly just coincidence.

To re-hash, the correlation in that data is not strong, maybe not even weak. You should gather much more data before even attempting to take anything from it. Some has been suggested (meat consumption in general, calorie consumption, activity levels, age, gender, etc). If you think that single graph is curious enough to raise the question, fine. But that means go on to these other areas, and see if anything holds. One chart is barely a starting point.

Perhaps it's just me and the world thinks differently but I have great difficulty in trying to see any positive results from that saying. I will take your advice to look little deeper.

 

[mfb]

The answer is not "NO" The answer is that no one really knows.

The question was "does this scatter chart show cause and effect", and assuming you mean "show" in the sense of evidence for it, the answer is no.
Even if there is cause and effect (that's what we do not know), the chart does not show it.

 

[OmCheeto]

This thread doesn't make much sense. Try plotting fatness vs. number of cars per person

Sense, schmense... This is fun!

Cars make you fatter than eating chickens makes you fatter!

, or just vs. GDP per capita. I bet you'll get a much better correlation.

I wish you hadn't asked that...

AFAIK chicken grow fast because they have ideal conditions. I know for a fact that temperature is regulated to $0.1^oC$ within an ideal temperature profile, probably also lighting does not resemble natural daylight.
Anyway, meat is meat, there are no homeopathic imprints or toxic substances in it.

It is probably true that developed countries eat too much of *any* meat, though. But I doubt statistics can shed some light on this, because there are too many other variables. You need a controlled experiment where the only difference is the amount of chicken eaten, over at least a decade, in a group of at least several tens of people.

Not sure where you're going to get a "control group".
But you are right, there are many variables.

For instance, working on the farm will make you skinny.

Anyone know how to add statistical data together, to get a really smooth graph?

I'm up to 15 variables now.

 

[Mark44]

Cars make you fatter than eating chickens makes you fatter!

What if you have two cars, and eat chicken in both of them?

 

[mfb]

For instance, working on the farm will make you skinny.

Or being skinny makes you sell your car and work on farms.

Your income/obesity outlier is Japan, by the way (same for the cars and agriculture), and you should consider GPD purchase-power adjusted.

 

[Mark44]

"correlation does not prove causality"

Perhaps it's just me and the world thinks differently but I have great difficulty in trying to see any positive results from that saying.

First off, that's not just a "saying." And second, yes, the world pretty much thinks differently from you about whether factors that are correlated mean that one of them causes the other. This is a result from many, many years of the disciplines of statistics and probability.

We can "agree to disagree" but you're not going to find anyone in your corner who is knowledgeable about statistics.