MAKING IT USEABLE    Interpreting Estimates       Limitations of NPTS

Examples illustrate the pitfalls of using unweighted data:

If it was desired to establish income quartiles from the NPTS data, the results from weighted and unweighted data would be:

Estimates of Income Ranges by Quartile
Income Quartile	Unweighted Estimate	Weighted Estimate
First	0 - $ 24,999	0 - $ 19,999
Second	$ 25,000 - $ 24,999	$ 20,000 - $ 34,999
Third	$ 40,000 - $ 59,999	$ 35,000 - $ 54,999
Fourth	$ 60,000 and above	$ 55,000 and above

 

Estimates of Persons per Household
	Unweighted Estimate	Weighted Estimate
Estimate of Average or Mean	2.65 persons	2.58 persons
Estimate of Median	2 persons	2 persons

These examples show that there is a bias in the use of unweighted numbers, with under-representation of one-person households, lower income households, and household with one or fewer licensed drivers

These biases become more marked when analyzing travel behaviors of subgroups of the population, which is a significant application of NPTS data

As has been stressed several times throughout this material, unweighted data should be used ONLY TO STUDY THE NATURE OF THE SAMPLE ITSELF AND TO REPORT ON THE SAMPLE

Also, to compute sampling errors for weighted data, the actual number of samples in the dataset (i.e., the unweighted count) is needed

Usually, the number of sample cases will be the only item of unweighted data that will be needed

For all other analysis of the data, WEIGHTED DATA SHOULD ALWAYS BE USED.

---