World Statistics Day was yesterday, but we’re celebrating all week long! This celebration means acknowledging the impact statistics has on our world. Who is your favorite statistician? Share with us who they are and why they top your favorites list.
Choose Language Hide Translation Bar
What Factors Affect Office Temperature? Data Collection and Cleaning

This post is the second in a series that will show how we conducted an observational study about office temperatures, from design to execution to analysis and visualization. For the first post, see Part I.

In the previous post, we discussed what factors might affect office temperatures. We used JMP’s design of experiments platform to design an observational study that would determine whether those factors are strong predictors. Today we’ll examine the process by which we collected the data and how the data was assembled for analysis.

We had nine digital thermometers on hand to help take temperature measurements, pictured at left. They were small, easy to carry from office to office. The metal sensor at the end of the wire was the part each volunteer needed to be careful of. The sensors were very sensitive to temperature changes, but often slow to change and stabilize. For example, if you accidentally touched the sensor with your hand, the temperature reading would rise quickly, but it would take several minutes to drop back down to room temperature.

We called for a new set of assistants each time we took a measurement, and got many different volunteers each time. Some people volunteered for most or all sessions, while others could only assist for one or two. Each volunteer was given a thermometer and an assignment for a sector. Remember, a sector has 10 offices. Each office has characteristics, such as whether it is on the east side or west side of the building, and what type of space it is. Volunteers filled in their assignment sheet, and turned it back in. Each form looked something like this:

Each volunteer had an assignment of 10 offices. Inside an office, they held the thermometer in the center of the room and waited a few minutes for the thermometer to stabilize. Volunteers also made note of the weather outside (rainy, sunny, etc.) and many jotted down notes regarding the conditions they found inside the office. For example, some offices were vacant with the lights turned off, while others had many lamps and multiple computers in the room. One room even had a space heater. A frequent comment was that rooms that were vacant and dark seemed to be consistently cooler than those that were occupied. These types of observations are not factors that we thought to include in the model, but this is one of the reasons this is called an observational study. If a more rigorous and controlled experiment were to be performed in the future, we would now know to also control for things like whether the lights were on.

Once the data was collected for all six recording periods, one lucky volunteer was assigned the task of entering data into JMP. Afterward, we used Tables->Concatenate to assemble data from all 54 “runs” into one large table. Once that was complete, there were a few maintenance tasks to perform on the data.

1. Create a new column named Fahrenheit that uses a formula to modify Y. Our measurements were taken in degrees Celsius, but I am personally much more familiar with the Fahrenheit scale, so I will use a Fahrenheit column in my analyses.

2. Add the Value Ordering column property to time of day. The default order in JMP is alphabetical, and “afternoon” will come before “morning” by default. I would like to change that so that in any visualization that we create, “morning” is listed before “afternoon.”

3. Run a distribution on Fahrenheit as a sanity check against errors in data entry. We see that the data in the first picture looks skewed by an outlier. Double-checking the data, I see that there was a typo during data entry. 242 was entered instead of 24.2. After correcting that, we see a distribution that is much more like what we expect to see.

Now I think we are ready to run our model and explore the data visually using JMP tools. What factors do you think will be significant in the model? In the next post, look for a sneak peek of new Graph Builder features in JMP 9.

What do you think of our process of data collection and data cleaning? What would you do differently?

Article Labels

    There are no labels assigned to this post.

1 Comment

What Factors Affect Office Temperature? Exploring What’s Significant - JMP Blog wrote:

[...] about office temperatures, from design to execution to analysis and visualization. See Part I, Part II and Part [...]