Warm or Cold Feelings: Estimating Biases in Field Studies

1. Motivation, specific objective: Lighting systems with LEDs that allows for control of illumination, in terms of the correlated colour temperature (CCT) and illuminance intensity, is currently being introduced widely in buildings resulting in decreased energy use for lighting systems. There is a great focus on reducing the overall energy use for buildings and the major energy-consuming factor in buildings is related to the regulation of thermal sensation. An earlier laboratory study with fewer biases showed that a shift in thermal sensation could occur when changing the lighting by altering the CCT. Using intelligent lighting control could therefore potentially benefit the overall energy use for buildings if being integrated in the thermal strategy. The overall objectives were to test if it is possible to detect an effect similar to the findings from the laboratory study on the thermal sensation when altering the CCT in real offices environments. The objective of this abstract is to discuss the experimental setup imposed in our field study where supplemental lighting is influenced by daylight and other strong biases. Differences that relate to biases should be estimated and accounted for, but the amount of controls are on the expenses of possible tests of the treatments in focus.
2. Methods: In a real office setting, we test the influence of three different CCT´s of 2800, 3100 and 4000 K on the workers thermal sensation and preferences. Three similar hallways and adjacent single or double offices at the Danish Technical University, Denmark (55°41'38.5"N 12°6'5.7"E) were used. These sections were equipped with LED lighting with central control of CCT and light levels. During the field study the light was on at all times starting from 7 am and onwards unless the workers purposely turned off the light. The electric lighting system was adjusted to 650lx at table height in all CCT settings and it was not possible to regulate the light intensity. The offices had windows to the outside, facing East or West. The experiment was conducted primo January 2018 and lasted for seven weeks. The period was carefully selected leaving minimal influence of the daylight as a central bias and to strengthen the influence of the electric light treatment. The settings of CCT (treatments) was changed each Sunday in a rotation between the 3 sections having 1 week in between being used as an anchor with CCT of 3100 K in all sections. In this abstract we focus on the anchor weeks and the deviation that relates to the different sections and people answering our questionnaires. A total of 48 workers were encouraged to participate in the study. The thermal sensation and visual preference were assessed by an online questionnaire with ten (10) specific questions sent out once a week. The workers were asked to rate their individual sensation of the lighting and thermal comfort while situated in their office.
Indoor climate measured as the temperature, RH and CO2 concentration in the offices during the experiment were logged every 5 minutes along with measurements of outdoor diffuse and direct radiation, temperature and humidity in order to exclude them as possible bias.
3. Results and discussion: This field study differed from (most) laboratory studies in its duration and number of biases. In real office settings, we investigate longer term effects of differences in supplemental lighting CCT on the workers. The workers and sections differed in various ways. Thus, the anchor weeks with CCT of 3100 K in all sections was imposed, to estimate the differences between sections, biases of dealing with different groups of people (sections) along with the influences of continuous changes in the outdoor radiation. The questionnaire lead to a relatively consistent response rate – resulting in between 28-31 answers the first three weeks and between 20-24 answers the last four weeks (out of 48 requests). The response rate on our questionnaire was in average around 50%, which partly reflect its’ length. A short questionnaire, opposed to a more extensive questionnaire, did not allow us to ask question, of e.g. work related distractions, tiredness, stress or sound levels, to further include and account for such biases in later findings, but we do not find that a lower response rate would have been advantageous. The differences being present between the sections were by anchor weeks measured once for each rotation and at the end. However, imposing the anchor weeks compromised the amount of answers in relations to the treatments of 2800 and 4000 K and extended the test period. Our weekly sent questionnaires (7 weeks) were answered by 19, 135 and 26 workers (180 in total), in relation to treatment settings of 2800, 3100 and 4000 K, respectively. If the study was conducted without the four anchor weeks the test period alone would have been 3 weeks. Execution within 3 weeks would have resulted in a higher response-rate and a more balanced number of answers in relation to the imposed
treatments. However, this would compromise the measure of differences that relates to biases influencing the treatment effects. The need of controls for biases will depend on the treatment strength.
4. Conclusions: Field experiments are characterized by having many biases and often a weaker response rate on questionnaire surveys. Our online questionnaire of 10 questions sent out once per week in a field study resulted in an average response rate around 50%. The response rate declined slightly after 3 weeks and varied between 42% and 65%. We are unsure of the need for imposed anchor weeks, since the need for controls of biases influencing the overall result will depend on the treatment strength and survey response rates. However, we stress that differences that relate to biases should be estimated and accounted for.