Six Reason to Question Vaccinations

[Beth]

Woah. So if crime increases in Gabon, and then the number of arrests increase in NYC, it implies a relationship ?

No, only if those events are correlated does it imply a relationship.

 

[JJM]

Those facts are common knowledge. You are claiming correlations exist. That's a very specific mathematical relationship. Get the data, compute the correlation coefficient and show that it's value is statistically significantly different from zero. Then I'll comment on the relationships.
{snip}

Yeah, sorry, asking you prepare an adequate response to that one was not fair. You don't know what "correlation" means.

 

[robinson]

Yeah, sorry, asking you prepare an adequate response to that one was not fair. You don't know what "correlation" means.

No, personal insults and being dumb are never required. In fact, being an ass just make you look bad.

 

[Beth]

Yeah, sorry, asking you prepare an adequate response to that one was not fair. You don't know what "correlation" means.

Based on your response, my assumption is that you have no idea what the actual correlation is between the variables you referenced and whether or not it's significant. Instead, you want me to buy your argument that they are, in fact, related and comment on their relationship assuming that fact. Then, when I refuse, you insult my understanding of the concept. That's a winning argument.

Incidently, the correlation coefficient (which is what is commonly referred to by "correlation") is the mathematical expression of the relationship between variables. It's computed by dividing the covariance of the two variables by the product of the square roots of their respective variances. It's a bit more complicated to compute the statistical significance of the correlation, but I can provide you with instructions on how to do so if you're actually interested. I doubt you are.

 

[T.A.M.]

While some may disagree, personally I think a large part of the reason the risk of those diseases is low is due to widespread vaccination.

It is a sound theory. That was my point. To not immunize ones children, is that not contributing to reversing this, or at the very least, is it not "riding for free" so to speak.

It's because MMR vaccine coverage in the West is 80-90%+, thus, if one individual does become infected, susceptibles do not interact with each other often enough to spread the disease.

ditto.

That doesn't apply for diseases such as Meningitis C or Hepatitis B though, which are pretty rare in the first place and usually require close contact with an infected individual to stand a reasonable chance of being passed on.

You are correct, that compared to MMR, Hepatitis, and to a degree, Meninigitis, are not as easily transmissible, but they are also much more serious illnesses, causing serious morbidity and mortality much more frequently.

Then you have the highly contagious but (in the vast majority of cases) mild diseases such as chickenpox. I doubt mass infant vaccination against it makes a blip in a graph of child mortality or morbidity.

I agree for the most part. While there is not doubt that Varicella is less lethal than some of the others, it can cause significant morbidity (40-50% of cases lead to at least one incident of the "pox" becoming superinfected with bacteria), and in rare cases mortality, with such sequelae as Pneumonitis, Meningitis. Rare though, for sure (much less than 1% of cases for meningitis and pneumonitis)

I am not a big fan of pushing the manditory use of the varicella vaccine, but I am not opposed to its use.

If a vaccine ever came out for the most common viruses causing the common cold (there are many), I wonder how many people would get it each year...a lot I am sure.

TAM

 

[JJM]

Your mind? That would be the only thing I can think of that connects the two. 28 cents a gallon and computer power increasing don't go together in time at all.

But they do correlate (inversely) with time. The prices just have no relationship.

Do you have a source of 28 cent gas, or $miilion PCs that are less capable than my pocket calculator?

They are occurring at the same time? What does this have to do with vaccinations?

Nothing; but see the argument in this thread by "Beth" who claims that correlation implies a relationship.

 

[JJM]

No, personal insults and being dumb are never required. In fact, being an ass just make you look bad.

Saying one does not understand correlation can be a fact, not an insult. Someone who says I don't know law is not insulting, it is a fact- I am not a lawyer.

Speaking of law, there is an old adage: "When you can't argue the law, argue the facts. When you can't argue the facts, argue the law. When you can't argue either, just argue." I see that here.

 

[fls]

Correlation means that two characteristics vary together. As Beth pointed out, two characteristics that vary because they are confounded (they vary together because they share an association with another characteristic) can still be reasonably said to be related. After all, we don't always know these things in advance.

However, if the variable that is shared is 'time', it becomes meaningless to call it a relationship. If not an outright tautology, the observation that variables vary with time is at the very least trivial. There are many characteristics that are progressive - the price of goods and services, the adoption of new technology, mortality rates, etc. The complaint levied (as far as I can tell) is that finding a significant correlation between two variables because both represent the progress of time is not a useful exercise.

Linda

 

[Skeptic Ginger]

What a wonderful example of selective quoting to attempt to look smart by making someone else look dumb, JJM.

We wear more Nikes in the US. Maybe that's why we have a higher infant mortality rate.

How about quoting everything Beth wrote on correlation:



The ironic thing that all of you have not noticed is that most of the correlations Beth highlighted supported your point of view. I.e., vaccination appears to be related to a drop in child mortality.

Are you really that uninformed about the scientific process? It has nothing to do with data supporting one point or another.

I am not so invested in my view of vaccines I wouldn't change it WITH SUFFICIENT CONTRADICTORY DATA. That's what you do with science, you follow the evidence. I posted the facts about the actual cause of the higher infant mortality rate in the US. It has been analyzed by many. THERE IS NO RELATIONSHIP TO VACCINES. Beth's claims stem from ignorance about correlations. Period.

But hey, Ivor, Beth and KellyB are the enemy, so let's ridicule them. How very...Human.

Running out of arguments or just getting testy?

 

[Skeptic Ginger]

Yes, it does imply a relationship. It just may not be a causal relationship.

No, it does not imply a relationship. Not in science. A coincidence is not a relationship.

Relationships in science can be associations. They don't have to be causal. But in this case, those are just two data sets, completely UNRELATED in any way except some imaginary one. And that is just as related as the examples JJM posted.

 

[Skeptic Ginger]

Show me the actual data and prove your claimed correlations actually exist and I'll speculate on what their relationship is. Correlation does imply a relationship, just not a causal one. For example, drownings increase with ice cream sales. The correlation is quite strong. The relationship is that both increase with temperature. They have a common causal factor.

That would be an association. There is no association between infant mortality rate and numbers of vaccines in the data you are claiming. There is a decrease in mortality of vaccine preventable diseases.

I can see you are missing the reason there is no association. I posted it twice now. I'll post it again.

If the infant mortality rate were consistent across the population in the US, there might be the association you claim. But it isn't. If you look at vaccinations and found the infant mortality rate really was inversely correlated, you would have a point. BUT THERE IS NO CORRELATION. The only correlation is when you lump all the infants in the US together.

I can't believe you are not aware of the error you are making here.


Try this. When we see some anecdotal evidence of something, we might choose to test an hypothesis based on the observations. But if we test the hypothesis and find out there is no association in the observations we thought we were seeing, we know there really was no association.

If you take the data on infant mortality, it is very easy to test if there is a correlation in infant mortality and numbers of vaccines given. When you try to correlate the data by numbers of vaccines, you do not see the pattern you are claiming. Why? Because the problem is race and income and disparity in health care, not numbers of vaccines!!!!!!!

 

[Skeptic Ginger]

Correlation means that two characteristics vary together. As Beth pointed out, two characteristics that vary because they are confounded (they vary together because they share an association with another characteristic) can still be reasonably said to be related. After all, we don't always know these things in advance.

However, if the variable that is shared is 'time', it becomes meaningless to call it a relationship. If not an outright tautology, the observation that variables vary with time is at the very least trivial. There are many characteristics that are progressive - the price of goods and services, the adoption of new technology, mortality rates, etc. The complaint levied (as far as I can tell) is that finding a significant correlation between two variables because both represent the progress of time is not a useful exercise.

Linda

There would be a correlation except in this case the numbers only appear to vary together because of the way Beth has chosen to count. If you accurately look at the data, the correlation she is claiming exists is not there.

 

[Skeptic Ginger]

I think you are the one who is mistaken, and I'm not trying to insult you either. Think about it. If a strong correlation does not imply a relationship, then what use is it? Why do we bother computing the numbers and testing whether a correlation exists?

In fact, any non-zero correlation implies that the two variables are not independent (i.e. they are related in some way) but in practice, small correlations are to be expected due to random chance alone. We can test correlations to see whether or not the meet the criteria to be considered statistically significant, but a general rule of thumb is a correlation of over .80 (or under -.80) is considered significant even for small sample sizes. There were strong inverse correlations between the measles vaccine and child mortality rates, though not for every country I tested. Why are you arguing that there is no relationship between the two.

Your examples are not representative of the problem. Your correlation of numbers of vaccines is not real. If it were real, you would see it when you controlled for other variables. When you control for income and race, your numbers of vaccines are shown to be no longer correlated.

 

[Skeptic Ginger]

No, personal insults and being dumb are never required. In fact, being an ass just make you look bad.

It's hard not to lose one's patience when such an obvious error in logic cannot be grasped.

 

[Ivor the Engineer]

It's hard not to lose one's patience when such an obvious error in logic cannot be grasped.

Though I find as time goes by, I'm getting better at keeping mine.

Let me spell it out for you:

Beth found an inverse relationship between mortality for children below 5 years-old and the increasing use of the DPT and Measles vaccinations in the US.

I.e. As the vaccination rate went up, deaths went down.

Do you really want to argue that this is not the case? Have I just been transported to wonderland?

With regards to correlation of variables, a classic example of a non-causal relationship is global warming vs. number of pirates. In this case, all the correlation can allow you to do is estimate the value of one variable, given the other.

However, when the two variables are strongly correlated and the outcome of the one variable has been shown to affect the other, as is the case for vaccination (why else are we bothering to vaccinate children in third-world countries if it doesn't?), it is not unreasonable to assume a common causal factor linking the two.

In the case of increasing vaccination rates being correlated with reduced child mortality in the US, one could speculate on reasons why this may be the case. For example, one explanation may be the increasing vaccination rate is a proxy for more parents being concerned with their children's health, thus providing them with better health care in general, as well as ensuring they have their vaccinations.

This site has some interesting graphs on infant and child health in the USA.

Here's a plot of infant mortality vs. time, and here's one of the leading causes of child mortality in 2004.

ETA: A plot of vaccine preventable diseases.

 

[fls]

There would be a correlation except in this case the numbers only appear to vary together because of the way Beth has chosen to count. If you accurately look at the data, the correlation she is claiming exists is not there.

You are simply talking about the difference between group-based data (in this case, the group is children in the U.S.) and individual data. It is not unreasonable to look for correlations using group data. It is done all the time. Comparing countries in terms of average alcohol consumption and rates of heart disease would be one example. However, as you point out, it is less accurate - especially for particular kinds of questions (associations with infant mortality being one). Which means that in some cases, even though one can go through the motions of applying the test and finding a correlation, one cannot conclude that there is an association. Especially if both variables are simply progressing over time (as in the case of infant mortality and vaccine uptake).

Linda

 

[Rolfe]

No, only if those events are correlated does it imply a relationship.

OK, JJM, you win. That's the most whopping fallacy/lack of understanding I've seen in a while.

Rolfe.

 

[Belz...]

No, only if those events are correlated does it imply a relationship.

And what does "correlation" mean to you ?

Correlation: the degree to which two or more attributes or measurements on the same group of elements show a tendency to vary together.

From Dictionary.com. The attributes may vary together, but does it mean they are linked ?

 

[Beth]

From Wikipedia: http://en.wikipedia.org/wiki/Correlation

In probability theory and statistics, correlation, also called correlation coefficient, indicates the strength and direction of a linear relationship between two random variables.

Do we really need to start a thread on the meaning of the term correlation? It has a very well defined meaning in statistics. It's a formula for computing a measure of the strength and direction of a linear relationship (or association if you prefer that term) between two variables. Any correlation coefficient statistically significantly different from zero indicates the two variable are related in some way and thus one variable can be used to make predictions of the other.

And Skeptigirl, I wasn't looking at infant mortality rates. I was looking at the mortality rates of children under 5.

 

[Hardenbergh]

One of my former co-workers made the decision not to have her children vaccinated. She home schools her children. She feels very strongly about this issue and I've read one of her letters on the editorial page of the newspaper on this subject.