1992 CENSUS OF TRANSPORTATION
1992 TRUCK INVENTORY AND USE SURVEY
=>SAMPLE DESIGN
The sample was drawn from a universe of active registrations in
each State at some date, generally between July 1 and December
31, 1992. If necessary an adjustment was made to account for new
truck registrations which occurred after this date but before the
end of 1992. The universe excluded those trucks that were
identified, from the registration information, as outside of the
scope of TIUS.
The trucks were selected using a stratified, random sample
design. The universe, or population, of trucks within the State
was divided into five strata: "pickup", "van", "single-unit
light", "single-unit heavy", and "truck tractor". The "pickup"
stratum consisted of all pickup trucks. The "van" stratum
consisted of panel trucks, vans (including minivans),
utility-type vehicles (including jeeps), and station wagons
on truck chassis. The "single-unit light" stratum consisted of
all single-unit trucks (excluding those in the pickup and van
strata) with a gross vehicle weight (GVW) of 26,000 pounds or
less. The "single-unit heavy" stratum consisted of the remaining
single-unit trucks. The "truck-tractor" stratum consisted of
only truck tractors. Within each of these strata, a
predetermined number of trucks were selected for the sample. All
trucks were selected at random with equal possibilities of
selection within a stratum.
=>SURVEY METHOD
For each selected truck, a report form was mailed to the owner
identified in the State's registration records. The owner was
asked to respond only for the truck identified by the vehicle
registration information imprinted on the form, regardless of
whether or not he still owned the vehicle. The information
received on the returned questionnaires was processed through an
extensive computer edit. Reports which contained questionable
responses were reviewed and corrected if necessary. Because
estimates are to cover calendar year 1992, if necessary an
adjustment was made to account for new trucks registered
after the date covered by the sampling frame but before December
31, 1992.
In each stratum, estimates of the number of trucks for each
characteristic were developed by expanding the observations from
the respondents to represent all trucks in the stratum within the
scope of the TIUS. The stratum estimates were then summed across
strata to form the estimates published for the State. This
methodology to account for trucks purchased new and registered in
the latter half of 1992 which were not covered by the sampling
frame differs from that used in the 1987 TIUS.
=>RELIABILITY OF THE ESTIMATES
The accuracy of the survey results is determined by the joint
effects of sampling variability and nonsampling errors. The
sources of error are discussed in the following paragraphs.
=> SAMPLING VARIABILITY
The particular sample drawn in the State is one of a large number
of all possible samples of the same size that could have been
selected using the same design. Estimates derived from these
different samples would differ from each other and from the
unknown total that would be obtained if all trucks in the State
were surveyed (the universe value). Ignoring the effects of
nonsampling error, the average of these estimates would equal the
universe value. The standard error of the estimate is a measure
of the variability among the estimates from all possible samples
of the same size and design. It measures how precisely we can
expect to estimate the unknown universe value. The relative
standard error (RSE), expressed as a percentage, is the standard
error of the estimate times 100 divided by the value being
estimated. Note that the RSE's given in table 2a are derived from
the sample and are themselves subject to sampling variability.
An estimate and its standard error, developed from a particular
sample, can be used to construct an interval estimate called a
confidence interval. (The standard error referred to here is
itself an estimate developed from the sample.) Associated with
each interval is a percentage of confidence (for example, 90
percent) which should be interpreted as follows. For each
possible sample, assume that an estimate and its standard error
were obtained. Then, for about 90 percent of all the samples, the
interval from 1.65 standard errors below to 1.65 standard errors
above the estimate would include the unknown value being
estimated. The following is an example of a confidence interval
calculation: Assume the number of basic platform trucks given
in table 2a is 20.5 thousand with an RSE of 10.2 percent. Then
the
standard error of the estimate is 20.5 x .102 = 2.09 thousand
trucks.
Now, the 90 percent confidence interval (the estimate plus or
minus 1.65 standard errors) is 20.5 plus or minus 3.4, or 17.1 to
23.9 thousand trucks. In table 2a, some data cells have RSE's
that are large, and the resulting confidence intervals could be
quite wide. The user should use such estimates with caution.
=> NONSAMPLING ERRORS
Nonsampling errors cover all sources of errors in the estimates
that cannot be attributed to sampling variability. This includes
errors in the reporting, collecting, and processing of data as
well as the inability to obtain responses from some sampled
units. Nonsampling errors lead to biases in the estimates. Bias
exists if an estimate, averaged over all possible samples, does
not equal the true value being estimated.
A major source of possible bias is nonresponse. There are two
types of nonresponse. "Total nonresponse" occurs when no
response to the questionnaire was received. In most cases, the
form was never returned to the Census Bureau, after several
attempts to elicit a response. For the United States,
approximately 90.2 percent of the questionnaires were returned
with some response. "Item nonresponse" applies to an
individual item or question which was unanswered, although some
response to the questionnaire was received. Several followups,
by mail and telephone, were done to reduce both types of
nonresponse. The details to account for total nonresponse and
item nonresponse are given below:
For most sections in table 2a (TI92US2A.dbf), total nonresponse
is handled, within the estimation procedure, by allocating
characteristics to the total nonrespondents in proportion to the
characteristics observed for the respondents. The amount of bias
introduced in this way depends on the extent that the
nonrespondents differ, characteristically, from the respondents.
For most sections in the table (TI92US2A.dbf), item nonresponse
is included as a separate line. For example, respondents who did
not indicate the major uses of their trucks are included in the
"not reported" category. This line shows the part of the total
estimate (for that table section) which is missing from the
estimates by major use. Users should exercise caution in
allocating the not reported figure to the major uses, since the
characteristics of item nonrespondents may differ significantly
from those of the respondents. For some questions, a response
was generated if it could be derived from other data. For
example, engine and body characteristics were frequently
determined through analysis of the vehicle identification number
and charts based on manufacturers' specifications. Missing length
and average weight data were imputed for each individual truck
based on the response from a record with similar characteristics
which were correlated with length and/or average weight.
=>APPROXIMATING UNPUBLISHED RELATIVE STANDARD ERRORS
The relative standard errors (RSE's) are presented for only the
row and column totals for U.S. tables 3 through 15 and 17.
The relative standard errors for an individual cell in U.S.
tables can be approximated by using the appropriate two-step
procedure below. Actual RSE's, estimated from the sample data,
are available on request.
U.S. tables 3 through 9:
First, calculate the standard deviation for the U.S. table
cell:
RSE(C) = RSE(S) * (SQRT((S(US-C))/(C(US-S))))
where:
S = the smaller of the number of trucks in the row and column
containing the cell
C = the number of trucks in the cell
US = the number of trucks in the United States
RSE(S) = the relative standard error corresponding to S
U.S. tables 10 through 15 and 17:
For tables 10 through 15, replace "number of trucks" by
"truck miles;" for table 17, replace "number of trucks" by
"number of trailers." Although either the row or column total can
be used, it is usually best to use the smallest of the two for
RCT.
Example:
Suppose, in U.S. table 3(TI92US3A.DBF), there are an estimated
67,000 trucks in the cell for agricultural multistops or
stepvans, for which an approximation of the RSE in percent is
needed. To approximate the RSE in percent for the agricultural
multistop or stepvan cell, the following information must be
extracted from the table (1) 59,200,800 trucks in the U.S., (2)
3,554,600 trucks and an estimated RSE of 2.3 percent for the
"Agriculture" column, and (3) 408,400 trucks and an estimated RSE
of 2.9 percent for the "Multistop or stepvan" row. Since the row
total of 408,400 is less than the column total of 3,554,600, use
the row figures to approximate the RSE in percent:
RSE(67,000) = 2.9 * (SQRT((408,400(59,200,800-67,000))/(67,000(59,200,800-408,400)))) = 22.7 percent
Some exceptions from this procedure will yield better
approximations of the relative standard error in particular
cells. Certain rows and columns in the tables are composed
predominately of trucks, excluding pickups and vans ("large
trucks"). Because of the sample design, one obtains a better
approximation of the relative standard error of the estimate for
a cell within a row (column) of "large trucks" by using the row
(column) total even though the column (row) total might be
smaller. When both totals consist of "large trucks," use the
smaller of the row or column totals.
Columns of predominately "large trucks:"
Tables 4 and 11--Light-heavy and Heavy-heavy
Table 5--50,000 to 74,999 miles and 75,000 miles or more
Tables 7 and 13--All except Single-unit 2 axle trucks
Rows of predominately "large trucks:"
Body Type--All except Pickup, Panel truck or Van, and Multistop
or Stepvan
Annual Miles--50,000 to 74,999 and 75,000 or more
Range of Operation--Long range (more than 500 miles)
Gross Weight--19,501 pounds and over
Lease Characteristics--Leased with driver
Hazardous Materials Carried--All carrying hazardous materials
Miles per Gallon--Less than 5 and 5 to 6.9
Equipment Type, Braking System--Air
Truck Type and Axle
Arrangement--All except Single- unit 2 axle trucks
Cab Type--All