MAKING IT USEABLE    Interpreting Estimates       Calculating Confidence Limits for Subsamples of the Data

Just as sampling errors were calculated for each different level of geography that was subsampled, so also confidence bounds can be calculated on the statistics for subsamples. Whenever a sampling error can be calculated, confidence bounds can be estimated.

It is useful to see how the confidence bounds change with reducing subsample size and scope by calculating the confidence bounds for various state and city subsamples.

Note:  The examples shown in the following table are for illustrative purposes only.  It is not appropriate to analyze NPTS data for specific states or metro areas.

(WARNING:  Failing to use confidence limits, derived from the sampling errors, would permit one to draw erroneous conclusions about differences in values between different geographic areas, where the samples are sufficiently small and the sampling error sufficiently large that no conclusion of difference can be supported. Simple numeric differences can, therefore, be very misleading and should not be used to draw conclusions about sample data.)

Table 11
Examples of Sampling Errors for Various State and City Subsamples for Household Size and Work Trips per Household

---