Cognitive Studies In Context
Laboratory Module
Effects of Cell Phone Use on Driving Performance: Investigating the Mechanism of Interference
Appendix A: Data Analysis
by Leon Rozenblit
Department of Psychology, Cornell University
lgr4@cornell.edu
So far
So far, you have run the study and have collected the data from each trial into a separate worksheet in the Driving Raw Data DB.
Getting a Performance Score for Each Trial
You will notice that
each worksheet contains a group of functions (outlined by a heavy box)
that compute a performance score from the raw data. Hint: You can use the Auditing/Trace Precedents command (found
under the Tools menu) to help visualize the relationship between the
formulas.
The Driving Raw Data DB workbook is designed to give you some
flexibility in analyzing the data from your experiments. Consequently,
you must determine and enter certain values into the worksheet to
compute the performance score.
To obtain a performance score you will
- Determine and enter the Optimal Run Time for the scenario.
- Determine and enter the appropriate coefficients for the computational formulas.
Optimal run time
Optimal run time is the minimum time
in seconds a competent driver would take to complete the run for a
given scenario while obeying all traffic rules and avoiding accidents.
To determine the Optimal run time
- Become familiar with the scenario (by driving through it several times).
- Measure the time, in seconds, it takes you to complete the
scenario without any mistakes, speed exceedances, or traffic
violations.
When you are done, enter the Optimal run time into the spreadsheet.
Hint: you will probably want to record the optimal run time for the scenario in your notes, for future reference.
Coefficients
Conceptually, the performance score is computed from raw data by treating each raw data item as a term in a weighted sum.
performance score = (k1)term1 + (k1)term2 + . . . + (kn)termn
You can change the weight of each term by changing its coefficient (kn).
The coefficients ought to reflect, to the extent possible in a
simulation, the real world motivations of a driving task. Thus,
mistakes are negatively weighted. Furthermore, serious mistakes should
be more heavily weighed than minor mistakes. You can modify the coefficients for each of the terms (except the
first one) by entering new values into the Coefficient column on the
spreadsheet.
Exception: Mean time to collision, deviation
"Mean time to
collision, deviation" measures how closely the driver came to colliding
with other cars. Small values mean the driver came very close to
colliding with other vehicles. The "deviation" refers to the standard
deviation of the distribution of all times to collision for the run.
The values may provide some indication of "how closely the driver is
cutting it."
To modify the coefficient for this term, click on the cell containing the formula and edit the formula in the Excel formula bar.
Standard settings for Coefficients
You can use the following settings as a starting point:
Mean time to collision = | 0.25 |
Off road accidents = | -80
|
Collisions = | -120 |
Pedestrians hit = | -160 |
Speed exceedances = | -20
|
Speeding tickets = | -20 |
Traffic light tickets = | -40 |
Stops at traffic lights = | 40
|
Length of run (seconds) = | 1 |
Analyzing the performance scores with StatView
First, you must
enter the performance scores into a StatView data-set. You may use the
"StatView Data-set base" file in the Driving Module folder. If you are
unfamiliar with StatView, use the StatView guide (under the balloon
help menu) to learn how to perform specific tasks. You should refer to
the manual for general advice on data analysis. Second, you will use one of the StatView templates to analyze the
data. Hint: you will probably use the paired t-test for preliminary
analyses in Exercise One.