Choosing the Ordering of Your Sample

You may also have noticed that the random numbers were sorted into ascending order. In most cases we want the program to do this because it will make our research easier. For example, marking those people to be interviewed on our employee list would go faster with the random numbers in order. However, there are situations in which you do not want your sample to be in order. That is why the program lets you choose, before you run the sample, what order you want the resulting sample in, by selecting c, for choices, in the main menu.

A case in which ordering the sample would be harmful is if the subjects had to be examined in random order. For instance, the investigator might worry that those subjects interviewed later might have talked with those interviewed earlier. Interviewing in random order would not prevent this, but it would make it easier to perform statistical tests to determine whether there was an interview-order effect. A similar problem confronts opinion surveys where the result may be affected by world events which happen while the study is in progress.

Another problem with ordering the sample occurs if the research has to be suspended before every individual selected in the sample is examined. The partial data is somewhat suspect since only part of the population had a chance to be selected. Suppose in our example above that the employee list was in order by seniority with the most senior employees having the lowest numbers. If the data were only partially collected the sample would be biased against newer employees. Or suppose that the employee list were in alphabetical order. In this case the bias is less certain, but we might worry that different ethnic groups have names that tend to fall in in different parts of the alphabet. Note that in this last case it might very hard to detect whether the sample was in fact biased except by another research study.

 Let us assume that we have decided to rerun our sample selection, to give us a sample in random order. Since we are at the main menu, we can select c for choices. The display then reads:

                              SET THE CHOICES
   The following options may be set or changed by typing the letter
   s -- sort the sample when it is drawn
                     the sample will be sorted
   p -- print the sample as it is generated.
                     the sample will not be printed.
   t -- save the sample to a file for use in other programs.
                  the sample will be saved in a file named unnamed result
   d -- draw a supplemental sample
                     a supplemental sample is not being drawn.
   r -- retain the raw sample in case a supplement is needed.
                   the raw sample will not be retained. 
   To quit this menu just press return.

As you can see from the choice menu, the sample was supposed to be sorted. If we want to change this, we just select s and answer no when the program asks us if we want to sort the file. Since in most random sampling studies the sample is sorted, the program always starts with this choice set to yes. We can now rerun our sample by pressing return and then selecting p for pick from the main menu.

We have now accounted for the principal features of any random sampling design. Determine a sample size, select random numbers in the appropriate range, and sort them. However, the sample just described is a very simple one. Many types of research will call for the drawing of more complex samples.


this page is at