> First, we analyzed 38,851 responses to the CBE Occupant Survey about satisfaction with temperature from 435 office buildings across 168 cities in the US > We

> First, we analyzed 38,851 responses to the CBE Occupant Survey about satisfaction with temperature from 435 office buildings across 168 cities in the US > We collected 16,791 tweets with common expressions of cold discomfort in U.S. offices between January 2010 and December 2019 > Our findings demonstrate No. The findings don't support the conclusion. Sorry. The conclusion is likely correct (read on), yet the methodology in this study is unlikely to actually support it. Statistical inference requires that the following be true: - The data/samples are randomized - The data/samples must be independent - The population distribution must be normal (or, at a minimum, reasonably symmetric about a peak) - Ideally, the population standard deviation is known Yes, there are techniques one can use to reach conclusions when some of these rules are violated. However, one must be highly skeptical of the results in such cases. I say "data/samples" because to engineers a sample usually means a single number or single data point (as in a sample from an analog to digital converter or a sample used in a finite impulse response filter); this is something that always twisted my brain around whenever I had to switch from EE/DSP work to thinking about statistics. In statistics as sample is a set of data points randomly chosen from a population. To pound this one to death, in this context "sample" refers to a bunch of data points, not a single data point. In fact, one would ideally want a bunch of randomly-chosen sample sets. For example, 50 samples each with 10,000 randomly-measured data points. I would also suggest work could be done to try to sample a randomized control group in order to isolate the variable being evaluated. Not sure this would be the correct control group...one idea could be to ask randomly-selected people at the beach for a hot/cold assessment at a range of temperatures. That tends to eliminate most clothing as a variable. In this study I can't see how the samples were random. At the very least, we don't know who did not answer the survey. It is possible that 95% of building occupants, regardless of gender, are just fine and have nothing to complain about. We don't know how many data points come from the same office, building or geographic areas. Furthermore, we I don't think we know the temperature setting for each office or building, the condition of the air conditioning system (Was it working correctly? How accurate was the temperature control?) or how close/far each person surveyed was from a window (which could make you hot or cold), the air conditioning exhaust (which is generally very cold or hot), etc. It is easy to understand that it might be sensible to ask something like "Do you sit within a 1 meter radius of the air conditioning exhaust?" and either exclude those answers or perhaps, if justified, use them as part of a control group of sorts. Did people answer in the morning, after breakfast, mid-day, after a heavy lunch, mid afternoon, end-of-day, in the moment or at home away from the office? What was their diet? A high protein diet will tend to make you feel warmer after eating. Etc. I don't think we know that the samples were independent either. If the samples contain a bunch of people from the same office, are their answers truly independent? Again, how about those who did not reply? Were respondents biased in any way? Would the sample mean and distribution be far different and more normal if everyone (or a much larger percentage) of the population had responded? Etc. Having said that, it is clear that people, regardless of gender, will perceive the temperature of the environment differently. As a personal note, when I am leaner I feel cold far more intensely than when I am less lean. The difference is almost digital, on/off. Either I am perfectly comfortable in the cold or I am at the edge of shivering. I interpret that as being caused by the relative thickness and distribution of the adipose (fat) layer. Which leads me to think about this study/conclusion in a different way, one which supports the conclusion that women tend to feel colder than men despite my belief that the methodology used cannot support that conclusion. Women are generally characterized by a higher distribution of adipose tissue in the gluteal-femoral region, while men accumulate fat in the visceral region. Hence the pear-shape for women and apple shape for men. [0] Fat is an insulator. If you have more fat around the central/abdominal region this means you have more insulation in an area that might be a large contributor to the feeling of being cold. Women have relatively less insulation in this area, which means it stands to reason that they might feel colder under the same conditions. From the article: "In humans and other mammals, body fat distribution is significantly different between males and females and is a strong predictor of disease risk [7•, 8]. Women have more subcutaneous adipose tissue (SAT), creating a “pear shape” distribution, while men have fat, predominantly distributed to the visceral adipose tissue (VAT) around the abdominal organs creating an “apple shape” body habitus [9–11]. Central/abdominal (subcutaneous upper body and visceral) fat deposition correlates with an increased susceptibility for metabolic complications, while gluteal–femoral (lower body) adipose tissue distribution is associated with reduced metabolic risk and may be protective against the adverse health effects of obesity in both sexes [12–14]." Of course, the other question is: This is a very simple concept in Physics. This is high school thermodynamics. Was a study really necessary? I hope it didn't cost much. [0] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6525964/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6525964/