The State of Wrong

Personal Lessons from a class I took during Spring Quarter, NAS 212 Community Development for Sovereignty and Autonomy, at UC Davis.

Correction: There are many states. I have experienced them all. From a Type 1 error like not seeing that someone financially wealthy is poor in other ways, to the more problematic Type II error like thinking that poverty has a universal definition (1). There are ways to mitigate these errors but not eliminate them entirely. Let’s walk through some of these assumptions that I no longer have confidence in making the claim as entirely ‘right.’ In this week’s blog, I explore the phenomenon of being wrong, and share how it is a lesson I learn time after time.


Chocolate Stats 101: Correlation vs. Causation

Question: Does eating more chocolate make me happier? In a recent conversation with a Gates scholar, the topic of chocolate inevitably came up. I had spent the morning eating a large amount of chocolate, which I presumed was giving me writing inspiration. Chocolate in my mind is associated with greater productivity, happiness, and overall level of wellbeing. However this Gates scholar made an interesting remark: the happier she is, the less chocolate she consumes. Let’s take a look at these two statements from a statistical point of view. 2015.1.23_ChocolateCorrelation1 Which graph is correct? This is a great example of the statistical pet peeve that many people believe correlation equals causation. Chocolate consumption is the independent or explanatory variable and Happiness level is the dependent or outcome variable; in other words, the level of happiness depends on the amount of chocolate consumed. The Line 1 (upward sloping) illustrates that eating more chocolate is associated with a higher level of happiness while the reverse is true in Line 2. Let’s take a closer look at Happiness as the outcome variable. This over-simplified graph only reveals just one explanatory factor for happiness—chocolate consumption—when in reality happiness is a much more complex and multi-dimensional construct. It is quite possible there are a number of “hidden” variables that influence happiness for that specific point in time (i.e. mood, physical, emotion, spiritual state). My colleague may eat more chocolate when she is eating depressed as a way to cheer up. The happier she is, the less she seeks this chocolaty outlet and her consumption goes down. This is an example of correlation: Chocolate consumption does not cause happiness but is rather a consequence of happy/unhappy states. On the other hand, I may eat chocolate because I think there is causal effect and the more chocolate I consume the happier I actually become. 2015.1.23_ChocolateCorrelation2 The next question is linearity. Suppose there is a causal effect between chocolate consumption and happiness. Does this hold true at all levels of consumption? Or beyond a certain threshold does eating more chocolate make me less happy? In economics, we use the term diminishing marginal utility to describe the incrementally smaller amount of utility (i.e. happiness) as our consumption of something increases. For example, the first chocolate square may bring euphoria. The second chocolate square tastes really good. And so does the third and fourth. But by the tenth chocolate square I am beginning to feel over-buzzed and guilty of over-gluttony. For each additional square of chocolate, I derive less pleasure from eating it. This would be a good example of a nonlinear relationship (upside down smile face). Another possibility is that I need some time to really feel the effects of chocolate. There is still an overall positive relationship, but between each spike in happiness there is a gap for the effects to sink in. So it’s not such a simple question of whether chocolate makes you happy. It depends on a host of other (hidden) factors, the direction, and shape of the relationship. Stay tuned for the next Chocolate and Statistics 101.