of different on of the main Impact of different building ventilation modes on occupant expectations of the main IEQ factors

This study explores the relationship between the perceived performance of specific IEQ factors and occupants’ overall satisfaction with their workspace. In particular we examine the influence of ventilation system type (i.e. Air-Conditioned AC, Mixed-Mode MM, Naturally Ventilated NV) on that relationship. Statistical analyses were conducted on the post-occupancy survey database from the University of California at Berkeley’s Center for Built Environment (CBE) to estimate the relative importance of individual IEQ factors on occupants’ overall satisfaction, depending on whether or not the occupants were satisfied with the IEQ factor in question. Based on these analyses, 15 IEQ factors were classified as Basic Factors , Bonus Factors or Proportional Factors, according to their relationship with overall satisfaction, as described in Kano’s satisfaction model. We found that the classification of some IEQ factors differed for the occupants of AC, MM and NV buildings, suggesting that occupants of buildings with different ventilation types have different expectations, and respond in different ways to various aspects of the indoor environment. A noticeable difference was in thermal environmental conditions: in NV buildings, good thermal conditions were associated with significantly enhanced overall satisfaction (i.e. strong positive impact), but there was little discernible adverse impact, even when thermal performance was deemed to be poor. In AC buildings, on the other hand, thermal conditions were more directly associated with negative overall evaluations of workspace by occupants (i.e. greater negative impact than positive impact). Finally, in MM buildings, thermal conditions exerted both positive and negative impacts of comparable intensities on overall satisfaction.


Introduction
It is not very difficult to find research literature comparing air-conditioned and naturally-ventilated buildings in terms of occupant comfort and the indoor environmental quality (IEQ) provided by those buildings. Occupants in naturallyventilated buildings are assumed to be more tolerant or forgiving of thermal conditions [1]. The thermal comfort zone In the following sections, an empirical test is conducted on this hypothesis: the relationship between perceived performance of specific IEQ factors and occupant overall satisfaction (i.e. Basic, Bonus and Proportional) can differ between occupants of buildings with different ventilation types (AC, MM and NV). This analysis is followed by a discussion of practical implications, focusing on how different occupant groups respond in different ways to various aspects of IEQ. The database also contains metadata describing technical aspects of the surveyed buildings, but not directly specifying the building's ventilation system. It does, however, contain information about HVAC systems, window airconditioners, use of natural ventilation, and presence of operable windows, which allowed us to infer the type of ventilation system in each building. Based on these inferences the survey dataset was sorted into three types of building ventilation system: Air-Conditioned (AC), Mixed-Mode (MM), and Naturally-Ventilated (NV) building groups. Samples with missing values on these metadata fields were excluded from our analysis, hence a total of 22,518 samples were retained for our analysis. In order to obtain some indication of the degree of Personal Environmental Control (PEC) between different building types, subjects with the various personal climate control methods were cross-tabulated with building ventilation type (

Identifying the impact of individual IEQ factors on overall satisfaction
In marketing research, a few analytical methods have been proposed to identify different categories of attributes (i.e.
Basic, Bonus and Proportional Factors) and tested for their validity. Dummy variable regression is regarded as a reliable, valid and practical strategy that properly accounts for the nonlinear relationships between attribute-level performance and overall satisfaction [8,9]. Consequently it became the most commonly used analytical approach in studies of the asymmetric impact of different attributes on overall satisfaction (e.g. [10][11][12] In order to estimate the differential significance of IEQ factors in association with their perceived performance level, survey samples were divided into three groups using dummy coding (coded 0 or 1); a) those who were highly satisfied with the IEQ factor in question (subjects who rated their satisfaction at the top two levels i.e. +3 and +2), and b) occupants who were highly dissatisfied with the IEQ factor (subjects who rated their satisfaction at the lowest 2 levels i.e. -3 and -2), and c) those occupants who were indifferent to the IEQ factor (subjects who rated their satisfaction level in the middle of the scale i.e. -1, 0, and +1). This third group is referred to as the reference group. The main reason for binning samples into three sub-groups (i.e. satisfied, dissatisfied and reference) lies in the uncertainty that the 7-point scale of satisfaction has the property of equal psychological intervals. For example, it is not verified that the psychological distance from a satisfaction vote of +2 to +1 is the same as the psychological distance from +1 to 0. Therefore we decided to simplify the 7-point scale responses by binning into three groups. The logic behind this binning is directly comparable to that used by Fanger [13] in his mapping from a 7-point scale of thermal sensation onto a thermal satisfaction/dissatisfaction bifurcation. He defined dissatisfied (the "D" in PPD) as those who vote -2 or -3, +2 or +3, based on the evidence from Gagge et al.
[14] -"real discomfort is first expressed by those voting higher than +2 or lower than -2".
Then multiple regression analysis was conducted with overall workspace satisfaction as the dependent variable, and the other 15 IEQ factors as independent dummy variables. Therefore, as defined in the equation (1), the regression analysis produce two coefficients for each of the IEQ factors: one for 'satisfied group' to measure the impact when performance of the IEQ factor is perceived as performing well, and the other for the 'dissatisfied group' to measure the impact when performance of the IEQ factor is rated as poor.
OS: occupants' Overall Satisfaction score with workspace b 0 : average of overall satisfaction score of reference groups Conversely, if the absolute value of b 2 on IEQ factor y outweighs that of b 1 , then factor y is grouped into the Basic Factor category. Finally, if the two coefficients for IEQ factor z have broadly the same absolute value, which means that both negative and positive impacts are approximately equal, then factor z is defined as a Proportional Factor. The procedure for the analysis is illustrated in Fig. 2  and NV buildings were rated significantly higher than AC buildings (p<0.05). Occupants of AC, MM and NV buildings were all close to neutral on the satisfaction/dissatisfaction scale for 'temperature', but MM building occupants were more satisfied with 'temperature' than their counterparts in AC buildings. MM buildings consistently achieved higher satisfaction ratings than AC buildings on most of the 15 IEQ factors. On the one hand NV buildings achieved the highest mean satisfaction rating for some IEQ factors such as 'amount of light', 'noise level', 'sound privacy', 'amount of space', 'visual privacy' and 'ease of interaction', on the other hand they were rated considerably lower on fit-out, cleanliness and maintenance issues (the six right-most IEQ factors in Fig. 3). Building age possibly contributed to low satisfaction scores on these IEQ factors. The majority of NV building occupants included in this analysis were in buildings constructed before 1960 (Fig. 4), whereas the sample of AC and MM buildings were spread more evenly across the time-period. So the furnishings may be perceived to be old-fashioned and fit-out materials could be deemed less clean in the older NV building stock.   Mean satisfaction scores by subjects with different means of personal environmental control are described in Fig. 5.

Fig. 4. Percentage of AC/NV/MM samples broken down by building completion year
Generally, occupants having access to both operable windows and HVAC controls are the most satisfied with their workspace environment (p<0.05), followed by those with access to operable windows (p<0.05), then individual HVAC controls (p<0.05), and last, those with no access to personal environmental controls at all (p<0.05).
Interestingly, none of the personal environmental control groups rated their thermal environment very highly (satisfaction ratings were all close to neutral and not significantly different between groups). Although occupants with access to both HVAC and operable windows had highest level of satisfaction with air quality, space, furniture, cleanliness and maintenance, the total size of this subgroup of occupants in the total sample was small (1.4% in Table   2).
Comparing occupants with access to operable windows with those who had individual HVAC controls, the former group seems to have the higher overall workspace satisfaction levels, as well as temperature, amount of light, visual comfort, noise level, sound privacy, amount of space, visual privacy, ease of interaction, comfort and adjustability of furnishings. The only IEQ factors on which satisfactions ratings for the operable window group were lower than the other control groups were related to cleanliness and maintenance issues (the three right-most IEQ factors), possibly due to ingress of external dirt and pollution.  were investigated and independence of predictors was established. Two regression coefficients per IEQ factor were derived from the procedure described in section 2.2; one to estimate the IEQ factor's impact on overall satisfaction when performance on that IEQ factor was deemed satisfactory, and the other when performance of the IEQ factor was deemed unsatisfactory.   their impact on overall satisfaction is relatively low (regression coefficient = +0.11: i.e. increasing the overall satisfaction score by 0.11). However when the AC buildings' occupants are dissatisfied with their buildings' thermal performance, the magnitude of the impact of 'temperature' doubled (regression coefficient = -0.22: i.e. decreasing the overall satisfaction score by 0.22). Thus, in AC buildings, the impact of 'temperature' on occupant overall satisfaction is bigger when the thermal performance of the building is perceived to be poor. Moreover the absolute magnitude of the impact is significantly different between satisfied and dissatisfied groups (i.e. the 95% confidence intervals of the positive and negative regression coefficients don't overlap). Therefore 'temperature' falls into the Basic Factor category in Kano's satisfaction model, so it can be regarded as a minimum requirement (expectation) for AC buildings, having a minor impact when performance meets expectations, but prompting significant overall displeasure when failing to meet those expectations. In MM buildings, the regression coefficients were the same with those of AC buildings, but the difference in the magnitude between positive and negative impacts was not statistically significant (i.e. the 95% confidence intervals of the positive and negative regression coefficients overlap). Therefore 'temperature' is classified as a Proportional Factor in MM buildings, exerting its impact on overall workspace satisfaction in approximately equal magnitude for both positive and negative performance. When a building's thermal condition is perceived as comfortable there is a positive improvement in the occupants' overall satisfaction with their workspace.
When thermal discomfort is experienced in an MM building, there is an equal but opposite effect on overall satisfaction rating. In contrast, 'temperature' in NV buildings had a strong positive impact on overall satisfaction (regression coefficient = +0.31), approximately three times bigger than that observed in AC and MM buildings.
Furthermore, the negative impact of 'temperature' on occupant overall satisfaction was statistically insignificant in NV buildings. Therefore 'temperature' fits Kano's definition of Bonus Factor in NV buildings -being forgiven [1] when perceived to be underperforming, but pleasantly surprising when performance exceeds expectations.

Occupants' different expectations for thermal comfort
In an earlier literature review on the relative importance of IEQ factors to overall satisfaction [15] we noted that most researchers ranked thermal comfort as the most important IEQ factor. But in the current analysis we found that thermal issues exerted their impact on overall satisfaction in different ways, depending on the type ventilation approach of the building. In Table 3, we noticed that 'temperature' was the only main IEQ factor classed into three different Kano's categories according to the ventilation type (except 'building maintenance'); 'Temperature' was classified as a Bonus Factor in NV buildings, but a Proportional Factor in MM buildings, and Basic Factor in AC buildings. In other words, provision of thermal comfort is apparently a minimum requirement for people working in centrally air-conditioned buildings. When the thermal performance of these buildings is deemed to be unsatisfactory, occupants' overall workplace satisfaction decreases significantly. In short, thermal performance inside AC buildings is likely to be noticed only when it fails to meet expectations. However, occupants of NV buildings apparently have relatively modest expectations of the thermal environment inside their buildings, and when temperature performance is good, exceeding their expectations such that their satisfaction generalises to overall workspace IEQ. Conversely, thermal discomfort in NV buildings doesn't necessarily translate to overall dissatisfaction with the building because it has already been factored into the occupant's expectations. Finally, 'temperature' effects on overall satisfaction in MM buildings seem to have characteristics that were identified in both AC and NV buildings. Occupants' overall workspace satisfaction level changes are commensurate with their perception of the building's thermal performance.
In other words, when a building provides a comfortable thermal environment, occupants will be satisfied and when the building is failing to deliver thermal comfort occupants will be dissatisfied with their overall workspace. This hypothesis for differential expectations of indoor thermal quality can be inferred from previous IEQ research literature: De Dear and Brager [16] argued that occupants in air- [18] speculation that thermal dissatisfaction doesn't necessarily result from non-neutral thermal condition, but rather when the thermal stimulus exceeds the limits of adaptive opportunity available to the occupants at that point in time.
When the cause of stimulus (like outdoor weather) is understood and good adaptive opportunity is provided, occupants' comfort zone is extended. To sum up, our analysis in this paper provides rare empirical evidence in support of a widespread belief that adaptive opportunity in the built environment influences the way occupants perceive IEQ.
In this analysis, buildings with the highest degree of adaptive opportunity, as assessed by personal environmental control items on the CBE questionnaire, were the naturally ventilated buildings, followed by mixed mode buildings, and then air conditioned buildings (see Table 2).
The analysis in this paper points to the possibility that occupants' expected or presumed levels of IEQ performance within a specific building might play a role not only in their reactions to thermal environment, but also the acoustic conditions. The regression analysis classified 'noise level' as a Basic Factor in AC and MM buildings, but as a Bonus Factor in NV buildings (Table 3). Perhaps this observation indicates that occupants of NV buildings are less sensitive to externally generated noise because they accept it as a necessary trade-off with thermal comfort in the operable window scenario.

Adaptive opportunity influencing occupants' satisfaction with various IEQ factors
Previous empirical studies usually link adaptive opportunity with occupant thermal comfort (e.g. [19,20]), but the connection between thermal adaptation and occupant satisfaction with other, non-thermal IEQ aspects is not common.
Baker and Standeven [18] suspected that restricted adaptive opportunity narrows the comfort zone and eventually heightens occupant sensitivity to other stimuli. In Fig. 5, we noticed that occupants with ample thermal adaptive opportunities (i.e. access to operable windows and HVAC controls) also expressed high levels of satisfaction with many of the other IEQ factors as well, compared to their counterparts in buildings with restricted thermal adaptive opportunities. Not wishing to overgeneralise this result, but it hints that adaptive opportunity is a generic attribute of buildings that influences not just thermal IEQ but other indoor environmental dimensions as well. This question deserves further investigation in future.

The effect of perceived control on occupants' attitudes and expectations
The psychology literature indicates that perceived control is a significant determinant of how we respond to aversive stimuli; our response to an event will be different if we feel that we were responsible for causing it, or whether it was the consequence of external forces beyond our control (e.g. [21,22]). The 'locus of control' theory in psychology, or perceived control over the indoor environment described as environmental empowerment by Vischer [23], suggests that a building occupant's attitude toward, and expectations of their building's thermal performance could be affected by where they perceive the 'locus of thermal control' to lie; internal versus external. In naturally-ventilated buildings occupants are fully responsible for achieving their own thermal comfort -internal locus of control. For example, people in NV buildings respond to a given thermal condition by availing themselves of adaptive opportunities such as operable windows, adjusting clothing insulation or drinking cold/hot beverages. Hence when they experience thermal discomfort, they tend to regard it as a consequence of their own response toward the thermal environment rather than the thermal environment provided by the building; we found thermal discomfort in NV buildings to be not significantly associated with negative overall workspace satisfaction (Table 3). In contrast, in centrally air-conditioned buildings occupants seem more likely to attribute thermal discomfort to their building's poor thermal performance rather than their own maladaptation to those conditions -external locus of control. They interpret their personal thermal experience as a consequence of external factors beyond their control (e.g. facilities manager) instead of ameliorating the situation through adaptive opportunities at their disposal. Our finding that thermal discomfort was significantly associated with negative overall workspace satisfaction in AC buildings (Table 3) is consistent with this interpretation.

Limitations of the study and suggestions for further research
Through a statistical analysis of CBE's post-occupancy survey database this paper identified differences in the impacts of indoor thermal environment on overall workspace satisfaction, depending on the building's mode of ventilation (AC, MM and NV buildings). We couched our explanation of these differences in terms of adaptive opportunity; with NV buildings having the highest degree of adaptive opportunity, AC buildings the least, and MM buildings somewhere in the middle. However, the meta-data in the CBE's database contains limited information regarding individual HVAC control and operable windows, so how effectively those features were in each sample building remains moot. Therefore the effectiveness or usability of personal environmental control in each building could not be reflected in our classification of buildings. When classifying buildings by ventilation type we inferred that presence of air-conditioning system and operable window corresponded to an MM building. But it is also conceivable that some buildings operating exclusively in central air-conditioning mode were classified as MM building because of the presence of operable windows, regardless of whether or not those windows were ever used for ventilation and thermal comfort. Anecdotal evidence from operators of many of Australia's recent MM buildings indicates they often revert to full-time AC mode a year or two post occupancy, for reasons that remain unclear at this stage. Others have noted that it is not unusual to find AC buildings with disused operable windows [24], and so misclassification of buildings between MM and AC categories in this analysis cannot be ruled out.
Another limitation is that the type of data available to this analysis was confined to just the subjective ratings. There were no instrumental observations of objective physical conditions to accompany the subjective POE data, so it remains unknown how those buildings in the database were actually performing. So while our analysis demonstrated that occupants of AC, MM and NV buildings all had different perceptions of their buildings' indoor environmental conditions, we cannot definitively say that this was the result of differences in actual physical conditions provided by the buildings. We acknowledge this as the main limitation of this study, therefore more detailed information about actual performance of each buildings and effectiveness or usability of personal control method in MM and NV buildings are crucial to make the result or discussions more concrete. This represents a fundamental limitation of contemporary POE techniques.

Conclusion
This study was based on the hypothesis that three types of relationship between occupant overall satisfaction and the because their evaluation of the overall performance of the building was not adversely affected. Also some other different patterns of occupants' responses toward various IEQ factors (such as 'noise level', 'ease of interaction' and 'building cleanliness') were detected, but we had insufficient building meta-data in the CBE database to make any inferences. To conclude, contextual differences seem to have an influence on occupants' satisfaction with indoor environmental quality.