In troduction
The relationship between symptom severity and lighting treatments was explored by means of Analysis of Variance (ANOVA). In addition, the Duncan multiple range comparison test was carried out to further evaluate significant differences between treatments.
Data was screened in the following manner:
• Data that did not list a participant, office or week number were deleted;
• Participants who worked in spaces other than those that had the experimental lighting treatments installed had their data during that period excluded;
• Participants who were identified as managers, pregnant or sick for a prolonged duration of the study were excluded.
Daily responses from participants were examined individually and as a group to isolate outliers or unusual data points that may have been incorrectly entered or were inconsistent with other responses (scatter plots). Single data points were only removed where an entry was inconsistent with the rest of the data set.
Several sets of analyses were carried out on the baseline and treatment data sets. These included:
• Screening symptoms to exclude those that did not reduce at the conclusion of the work shift;
• Examining the effect of a washout period (the first week of each lighting treatment removed);
• Examining the effect of age, gender and office differences on the study outcomes; • Examining only the responses in which symptoms were experienced.
These analyses were selected to determine if these factors influenced the study
findings and were in line with the exploratory nature of this study. The selected analyses were based upon previous research fmdings (as discussed in the L iterature Review).
The following sections outline the data sets tested. Each data set was tested independently, as screening the data sets reduced the number of data points available for the analyses and hence the accuracy of estimates of treatment effects. Testing combinations of these factors would have been desirable, for example treating baseline, age and gender together as covariates, but this screening would reduce the size of the data set to non-usable levels.
Baseline Data:
The data from the baseline period was initially examined to determine if there were significant differences between the participants in the three offices. If the three offices differed, then the treatment effects should be adjusted for these to remove any bias in the data. The data was tested with a one week and a three week washout period. Although a one week washout period should be adequate to enable any carry over effect to abate, the three week washout period examined the data when the majority of participants had been recruited and were familiar with the study questionnaire. The data set with the one week washout period was larger as not all participants completed all four weeks of the questionnaire. The complete and screened (symptoms that did not disappear outside of working hours deleted) data sets were considered for each of the analyses.
In this study, the baseline data did not represent a perfectly level 'playing field' for several reasons:
• Although all three offices had the same lamps installed for the baseline period, and the lighting was balanced as much as possible across the three offices, there were still some differences in the illuminance levels and luminance distribution in the offices;
• A lamp was removed from a luminaire in Office Two during the baseline period at the request of office management;
• Additional workstations were added to Office One during the baseline period, as
The following data sets were tested using ANOV A:
1. BW4 Baseline data, week four only (three week washout);
2. BSW4 Baseline screened data3
\
week four only;3. BW2-4 Baseline data, weeks 2-4 (one week washout);
4. BSW2-4 Baseline screened data, weeks 2-4.
Treatment Data:
The following data sets were considered for the lighting treatment analyses.
1. C Complete data set.
The complete data set included all four weeks of data within each trial period, and therefore had the largest number of participants and data points.
This data set did not include a washout period. Estimates of the size of the carryover effect, if present, were not possible to estimate accurately and therefore a one week washout was imposed for all other data sets.
2. cw 3. SW
Complete data set with washout. Screened data set with washout.
Data sets CW and SW included a one week washout period, enabling participants
to adjust to the new lighting treatment and allowing any carryover effect from the previous lighting treatment to abate.
4. CW-CVB
5. SW-CVB
Complete data set with washout - covariate baseline Screened data set with washout - covariate baseline
As previously outlined, differences between the participants in the three offices may introduce a bias to the data that may lead to inaccurate estimates of treatment effects (Ott, 1 993). In these analyses, the baseline data and thus the treatments were adjusted for any pre-existing differences between offices. The baseline data
used in these analyses were data set BW2-4 and BSW2-4, the average symptom
severity data from weeks two, three and four of the baseline period.
6. CW-CVA
7. CW-CVG
Complete data set with washout - covariate age Complete data set with washout - covariate gender
In these analyses age and gender were treated as covariates. Research literature has shown that the age and gender of participants may influence the incidence and severity of symptoms experienced (Skov et al., 1 989; Hedge, 1 990). Therefore this information was treated as a covariate in the data to determine the effect, if any, of these factors.
8. C-S
9. CW-S
10. SW-S
Complete data set - symptoms only
Complete data set with washout - symptoms only Screened data set with washout - symptoms only
These analyses only included participant responses when symptoms were experienced. Therefore this data focused upon differences in symptom severity, not incidence.
This data set considered those participants who may have been more sensitive to environmental conditions in the workplace.