Selecting a Simple Numerical Sample

In order to give us a concrete example to work with, let us invent a simple case. Suppose that we want to interview 60 workers at a business which employes 2,000 people. To make things very simple, let us suppose that we have a completely accurate employee list with the employees numbered from 1 to 2,000. In the absence of a computer we could take 2,000 identical slips of paper, number them from 1 to 2,000, stir them thoroughly in a large hat, close our eyes, select 60 slips, and interview the people whose names corresponded to those numbers. But the computer allows us to do this with less trouble, and in a way that is easier to report on.

To start, select d (for describe the sample) from the main menu. If you are not at the main menu, press return to take you back to it. The display will say:

    I'll ask you for some information about how to draw your sample
   What is the name of the most general category?

In this case we have only one category and it makes sense to call it employee. The machine will respond:

   Is employee numbered or named?

Now of course we know that our employees have names, but remember that for this sample we decided to draw numbers and then use those numbers to pick the names off of the list. We often use numbers this way when we do computer sampling. The alternative would be to tell the computer the names of all the employees. We respond with the answer numbered.

   What is the lowest number for employee?

Since we decided to number the slips to be drawn starting with one we type 1. The program responds:

   What is the highest number for employee?

and we type 2000 (in this program we never use commas or decimal points in numbers.) The program asks:

What is the name of the next most general category?
     If you are done just press Return

and since we are done we press return and the program displays the design that we have specified. In this case the design takes only one line.

   Category 1 -- employee numbered from 1 to 2000

The program now automatically returns to the main menu. We select p to pick the sample and respond to the question.

   What size sample do you want drawn?  You calculated 272

Since for this example we want a sample of sixty, we type 60. There will be a pause for selecting the sample during which the program reports, in steps of ten, how many numbers it has drawn.

Drawing the sample.
I'll count the number drawn.

Then a list of sixty numbers, all between one and two thousand appears scrolling on the screen.

employee	370	
employee	404	
employee	459	
employee	769	
employee	780	
employee	1126	
employee	1185	
employee	1221	
employee	1328	
employee	1372	
employee	1385	
employee	1988	

Most of the numbers will scroll irretrievably off of the top of the screen. But if you used "Choices" from the main menu and then option "t" to save the results, the program is actually saving them all.

In our example, we could now go down the employee list and interview those whose number appeared on the random list. If you tried to run this sample on the program, you may have noticed that the list of sixty never repeated the same random number twice. One assumption of random sampling designs is sampling without replacement which means that, once chosen, an individual should not have a chance to be selected again. The program automatically sees to it that you do not select the same item twice.


this page is at