LETTERS: STATISTICAL SAMPLING IS WIDELY USED

To the editor:

I would like to discuss some issues regarding statistical sampling in response to your Aug. 29 editorial, "Census 2000: Abide by the Constitution." I will not address any of the political implications you discussed in your editorial. I will address only technical issues regarding statistical sampling.

In your editorial, you state, "The Clinton administration wants to replace actual counting of individuals with a form of statistical sampling. This is wrong, even laughable when you think about it." Are you implying that any use of statistical sampling is wrong and laughable? I hope not, because today our society uses statistical sampling to make countless very important decisions. Here are some examples:

--The very rigorous testing of practically all pharmaceutical products is done using statistical sampling.

--The advertising industry uses the Nielsen ratings to make multimillion-dollar decisions regarding the value of TV commercials based on statistical samples of about 1,800 families.

--Businesses use statistical sampling to determine the quality of their processes and products and to determine customer satisfaction.

--Countless surveys regarding educational, social or political issues are published routinely (your newspaper does its fair share).

--Many organizations -- A.C. Nielsen, Harris Poll, political consulting firms, news organizations -- are in the business of conducting surveys using statistical sampling.

Why are there so many organizations willing to make substantial investments based on results from statistical samples? The answer is very simple. Statistical sampling, when done scientifically, provides very accurate results. Results from statistical samples are used to make estimates or generalizations about a population. Why not always count, measure or survey the entire population? Cost, time, feasibility and scope of information are among the several reasons.

Let's look at the issue of accuracy. Any time information is collected, there are several possible sources of error such as lack of access to some or many individuals or units, inconsistencies among individuals collecting data, measuring errors, incorrect reading of data, transposition of numbers, entering data incorrectly on a computer, glitches in computer programs and unethical practices by data collectors. Granted, the complete counting or measurement of the entire population would give the ultimate accuracy if done without errors. However, this is practically impossible, especially when a population is very large. Statistical samples can minimize errors through much tighter control of the data-gathering process. And such samples, although in most cases they are much smaller than the population, provide very accurate results. The most important factor that affects accuracy is the size of the sample. The larger the sample, the better the accuracy.

Using statistical samples of a relatively small size will provide very accurate results and, in most cases, more accurate than if one attempted to survey or measure the entire population, especially if the population is very large. Therefore, political issues aside, using statistical samples in doing the U.S. census not only is not wrong or laughable, it could provide very accurate results at a much lower cost.

Some of the information in this letter was abstracted from the book, "Business Statistics for Quality and Productivity," by D. Levine, P Ramsey and M. Berenson.

DEMETRIUS KARATHANOS

Cape Girardeau