Your experiment must be in a controlled environment and eliminate as many variables as possible, including anticipation of your test subjects.  More importantly, there must be a "control" where all things remain as they were.

This will be a difficult experiment to quantify since you must eliminate such variables as decorating taste and it must be of such a duration  that your results will be definitive.

Next, you must formulate your hypothesis based on known science and make your predictions. 

Next, after all your results have been tabulated, we will falsify your conclusions to see if the result is truly valid.  Even if found valid, later review may determine that your falsified conclusion was unsound, meaning it has been eliminated because of technical flaws in the devising of your experiment.

1. Develop a baseline of activity for your test subjects according to color of the room as it exists.  What was the average age of persons in the cafeteria?  How many people twitch as they sit down to eat?  What music (if any) was playing?  What was the temperature and aroma of the cafeteria?  What foods sold the best, indicating mood of those in the cafeteria?  Owing to seasonal variations, and other possibly unpredictable factors, your control study would require at least a year to complete.
2. Change one variable, say the color of paintings on the walls.  They should be the same subject matter but change only the color.  What is the twitch rate?  What caused the twitching?  Is it a significant result?  That will take another year.
3. Return the old paintings to the walls but re-paint the walls to another color.  What is the twitch rate?  What foods did your subjects select, etc?
4. Compile your data.  You may well find that there was no significant change, which will be an important result: that color is not a significant factor in decorating, as long as it isn't just plain disgusting.