• Second Stage: A six-hour time interval was selected between 6 a.m. and midnight during which the
researchers in a selected PSU monitored the EMS/police radio frequencies to be able to reach the crash
scene before it was cleared. The selection at this stage was done according to the sampling procedure,
“systematic probability proportional to size,” where the size was the number of crashes that occurred
during the same time interval in the previous year as coded in NASS-CDS.
• Third Stage:
Each week four days were selected from the selected time interval during which the
researchers in the PSU responded to crashes. This selection used the same procedure as the second
stage, although the size was the number of crashes that occurred on the same day in the previous year
as coded in NASS-CDS.
• Fourth Stage:
A crash was selected. Once a time interval and day of the week combination (i.e., time
block) was selected, the researcher responded to every crash during this time block until the first
qualifying crash for NMVCCS occurred.
In order to handle special situations, such as larger volume of transmissions or very large geographical
area in a PSU, sub-sampling was implemented in certain PSUs. In addition, due to operational issues,
some other local adjustments such as longer or shorter time blocks, etc. were made in certain PSUs. For
a more detailed description of the sampling and weighting procedure used in NMVCCS, refer to the
NHTSA technical report
1
on the NMVCCS sample design (DOT HS 810 930).
Sampling Weights
Each NMVCCS case has a value in the variable CASEWGT (“Case Weight”). This variable gives the
statistical sampling weight of the case for national representation. CASEWGT should be used as the
weighting variable in all statistical analyses. The NMVCCS case weights sum to 2,188,970, the estimated
total number of NMVCCS qualifying crashes from July 3, 2005, to December 31, 2007.
The weights were derived through the procedure consisting of mainly two phases, the design weight and
its appropriate adjustment. The design weight is calculated by taking the reciprocal of the probability of
inclusion of a crash, which is the product of the selection probabilities at all stages of the sampling
procedure. The design weights are further adjusted to compensate for the crashes that could not be
collected due to operational issues and other challenges. As a result, the design weights of time blocks
with missing crashes are distributed to other time blocks that have a sampled crash in the same PSU.
Cases with Weight Zero
Of the 6,949 cases in the NMVCCS files, 5,470 cases have weights greater than zero. These 5,470
cases are the nationally representative sample of NMVCCS between July 3, 2005 and December 31,
2007. The remaining 1,479 cases have weights of zero. Zero-weight cases have not been removed
from the NMVCCS database so they are suitable for clinical, case-by-case evaluation. However, they
should not be included in statistical analyses for national representation. Note that in any software
package with a weight statement, using the weighted data will result in omission of any cases with weight
zero. Zero-weight cases fall into the following categories:
Cases Occurring Before July 3, 2005
832 cases were investigated during the transition period from January 1, 2005 to July 2, 2005, when the
data collection effort was being phased in.
1
Choi, Eun-Ha, et. al. (2008) A Sampling Design Used in the National Motor Vehicle Crash
Causation Survey, DOT HS 810 930, April 2008. Washington, DC: National Highway Traffic
Safety Administration.
NMVCCS Analytical User's Manual NMVCCS Sampling and Analytical Issues • 8