It’s been a month that we have been on lockdown in New York, so I was finally pushed to do something I’ve been wanting to do: use statistical DOE to profile how the different variables influence coffee taste. If you drink enough coffee, you know that there are several variables. And it can get even more complicated when you consider interaction between variables. In other words, there may be synergistic effects in play.

Experimental setup

For this experiment I am focused on three variables:

1. water/coffee ratio
2. bean brand
3. temperature

All batches were made using a typical drip machine that uses coffee filters. There were other factors I wanted to consider, but could not due to limitations:

1. grind coarseness
2. Brew method

The following were the variables and their levels:

variable	low level	high level
tbsp coffee per cup water	1 tbsp	2 tbsp
Coffee brand	Allegro	Breakfast blend
Temperature setting	regular	bold

I used JMP to build the design. It was recommending 12 runs, but I did not want to overwhelm the participants with too many tastes to keep track of, so I chose 7 runs.

I had two participants, whom I will call Janet and Alex (real names witheld for privacy). I asked them to grade each cup on a scale of 1 to 5, with 5 being the most favorable. I told them to keep the scale relative, meaning they should grade the best cup in the batch as 5, and the worst cup as 1.

Results

The model is used is a linear regression, which includes 2nd order interaction effects. Here are the parameter estimates for Janet and Alex:

Janet

variable	Parameter estimate
tbsp coffee per cup water	0.5
Brand	-1.5
Temperature setting	very small
Interaction of coffee/water and brand	-0.25
Interaction of coffee/water and temperature setting	-0.75
Interaction of brand and temperature setting	0.25

Alex

variable	Parameter estimate
tbsp coffee per cup water	0.625
Brand	-0.375
Temperature setting	-0.25
Interaction of coffee/water and brand	very small
Interaction of coffee/water and temperature setting	0.125
Interaction of brand and temperature setting	0.375

Conclusion

Janet appears to be more particular about her coffee. She prefers a lighter concentration, and is very sensitive to the brand. On the other hand, Alex is all about how hard the coffee hits you- the stronger, the better. It is interesting to note that the temperature setting has little effect by itself, but may amplify the effects of concentration for Janet or of brand for Alex. Nevertheless, the interaction effects are not as as trong as the primary effects.