Let's look at an example of poor data analysis:
Oh, parabens... a prime victim of the clean beauty revolution.
Parabens, highly effective preservatives often found on 'dirty' lists, owe their vilification to a study from about 10 years ago, with, you guessed it, poor data analysis. The study concluded that parabens were hormonal or endocrine disruptors and would increase breast cancer risk.
How did they come to that conclusion? By injecting rats with parabens.
But here's the thing:
1. Humans are not rats.
2. Injecting something into the bloodstream is not the same as applying it topically.
3. The amount injected into the rat far exceeded any amount someone would realistically be exposed to.
So, knowing that — did the study accurately measure what it intended to measure?
No. They intended to measure if topical application of parabens would cause endocrine disruption in humans, but the study tested injected application in rats.
Did they generalize the findings beyond the study sample?
Absolutely. Based on their findings with high doses of parabens injected into rats, they generalized their findings to topical applications of low doses of parabens in humans.