Introduction 

The Proof-of-Concept exercise had left us with an impression that creating a simulative environment involves an extensive amount of resources including time, brain power (extremely cautious and picky decision making), and expenses.  This exercise offers an alternative solution for dealing with situations that may not offer us the luxury to be perfectionists.    

The two hour exercise involved two major steps - a quick model of Cris's office was made by groups during class time as well as individually at home.  Another class period was spent recording qualitative (interior lux readings at various points) and quantitative (class survey on the effectiveness of the model) data of the models.  All thirteen models were made with simple building materials and with minimal attention to detail since no more than 4man-hours (1 man-hour = 1 hour of work by one person) were spent on each model.      

above, right: in-class exercise
above, left: the models are set up outside Cris's office for measurements.

The following references will be useful while interpreting the graphs and the data. Click on peoples' names to compare their models' interior views with the real space.

M1: Elena (2hours)
M2: Inga (2hrs, 30min)
M3: Sylvia (2 hours)
M4: Karen (2hrs, 20 min)
M5: Naoya (2 hrs, 30 min)
M6: Dror (2hrs, 5min)
M7: Jane (2 hrs, 20min)
M8: Gwelen (2hrs, 15min)
M9: Corey (2 hours)
M10: Eddie, Inga, Dror, & Elena (one class session-approx. 1 hr)
M11: Naoya, Manuel, & Corey (one class session)
M12: Gwelen, Sylvia, Jane, & Karen (one class session)
M13: Eddie (2 hours)
Real: Real space

Glass Transmittance: 2496/2610= 96%
 
Unfortunately, the information contained on this page is not complete.  We are still missing a checklist of items present in each model.  If this information was available, it would have been interesting to observe whether the quantity of items present has any correlation to the qualitative scoring of the models.

above, right: Gwelen and Sylvia during class exercise

Luminance Measurements

Two measurements were made for each model:

Reading A: by the window
Reading B: by the door

After adjusting the model readings with the glass transmittance value, the luminance values were expressed as a percentage value of outside conditions at the time.  The outdoor conditions fluctuated significantly towards the end of the readings, which will inevitably lead to higher error possibilities. 

Model/Real Space Readings (glass transmittance not adjusted):

Exterior Readings: 

Interior/Exterior ratio (glass transmittance adjusted): 

The final plot and the graph reflects the % error between the real space inside/outside ratio vs. that of the model space.  The luminance ratios were generally more accurate and less varied near the windows. The average error for the A readings was 34% (std dev: 35%), while it was 135% (std dev: 199% !!!) for the B readings.

The following is a closer look at the chart (maximum value set to 200%):
B% Error for M7 and M8 can be considered as outliers.  When those two points were ignored, the average error for the B readings turned out to be just 65% (std dev: 65%) after all.

Time vs. Luminance % error

This was an attempt to explore the correlation between time spent making the model and the accuracy of the light readings.  In the following graph, time is plotted against the % error for the luminance readings.  Time is represented as % of additional man-hours spent in excess to the suggested total work time of 2 hours.
 
For example, if 3 people spent 1 hour each on a model, 

time=((1manhour/person * 3person)-2manhour)/1hour*100%=100%   

Once again, here's a closer look (maximum value set to 200%).  By observation, there is no distinct correlation between the time spent and the accuracy of the model.  The validity of this statement is questionable, since the conditions of making the models were not consistent (different settings, model makers exhibiting different skill levels, and inconsistent concentration levels while fabricating the models, for instance).  There is also no justice in judging the models only through quantitative measurements. 

Accuracy and Believability Survey 

In addition to the qualitative luminance measurements, the class also scored each of the models in two categories:

A: Accuracy
B: Believability

Note that the bold fonts signify a distinction from data points A and B from previous.

The scoring was based on a scale of 1 through 10, 10 being the highest and representing the qualities of the real space.  The following are the average values and ranks of the two categories (real space omitted from ranking):

The relative scores of the two categories (and consequently the rankings) were linearly related (almost 1 to 1 relationship), which can be seen from the following graphs:

Accuracy and Believability, Adjusted 

The following is an attempt to adjust the variances among individual responses to the survey.  For example, politeness was a factor.  While certain individuals felt reluctant to assign scores less than 7, others' scores ranged from as low as a 2.  Typical average score assignment of each person ranged from 5.5 to 8.1, while the standard deviation (variation of scores) ranged from 0.8 to 2.6.  Although this does not affect the general "ranking" of the models in an average sense, the variations of scoring of individuals invalidates the quantitative differences between each model's scores (in other words, the true score range of each person was not 1 to 10).  

By normalizing everyone's score range with a hypothetical average (7 was picked) and standard deviation (3 was picked), a better distribution of the scores can be observed (notice that the real space is no longer worth an average of 10 points).

Time vs. Accuracy and Believability

Once again, as an attempt to explore the correlation between time spent making the model and the qualitative effects it achieved, time is plotted against the A and B rankings. To overlay the values in perspective, time is represented as whole 5 additional man-minutes spent in excess to the suggested total work time of 2 hours (24-5 minutes).
 
For example, if 3 people spent 1 hour each on a model, 

time=((12*5-minutes/person *3person) - 24 * 5-minutes)=12 

Once again, a closer look (max. value set to 15):

A and B reading vs Accuracy and Believability

This is one attempt to see if your qualitative assessment of the space is paralleled by the quantitative.  A negative linear relationship of the graphs is sufficient to suggest that the two qualities maybe interdependent of each other.  Otherwise, this is just another method that proves that quality and quantity are distinct characteristics.  

A and B points are represented by their respective error percentage.  For ease of reading the data, the two outliers for point B readings have been omitted.  The data points for A and B are represented by ratios (fraction) between normalized average model scores and normalized average real space scores.

As seen from randomly scattered dots, there is no strong correlation between the quantitative readings and qualitative survey.  A mathematically calculated line for each of the graphs, however, proves that the scattered points approximately follow a negatively sloped line (for greater % error in quantitative reading, we generally give out lesser accuracy and believability points).  

Conclusive Remarks

The lesson we learned from this exercise is that it is possible to create 2 hour models that may significantly resemble the qualities of the real space.  This process involves prioritizing the key elements and making quick insightful judgments.    

My attempt in this summary was to put the data we have gathered as a class together and present them in different combinations.  A small amount of data can produce endless amount of graphs that may imply specific patterns. Any interpretation from these graphs (or decision to support of refute relationships derived from the data available) should be an individual choice.
right: a model in assembly

[  SYLLABUS | HANDOUTS | EXERCISES | STUDENTS | DISCUSSION | LINKS | WHAT's NEW ]

This WWW sIte is a class resource for the Spring 2002 session 
of Arch. 245: Daylighting in the Department of Architecture at UC Berkeley
© UC Regents 2002   Updated: Tuesday, February 24, 2004

Comments to Cris Benton at crisp@socrates.berkeley.edu
URL: http://www2.arch.ced.berkeley.edu/courses/arch245/Students/2002/Rosie_Chae/twohour/twohour.htm