Introduction to a variety of useful statistical ideas and techniques - Improving the User Experience through Practical Data Analytics - page 29

SIDEBAR: HOW TO DISPLAY CONFIDENCE INTERVALS ON A BAR CHART
IN SPSS—cont'd
6. If needed, scroll down to the bottom of the viewer to see the bar chart with the conidence levels
displayed:
7. To tweak any aspect of the chart, double click directly on the chart, right click and select “Edit
content in separate window.” This will bring you to the chart editor.
SIDEBAR: DOES ACTUAL POPULATION SIZE MATTER? IT MIGHT!
CONSIDER THE NEED FOR THE FINITE POPULATION MULTIPLIER (FPM)
When inding a conidence interval for a mean using Excel or SPSS (or for a proportion, covered
in a subsequent chapter), there is an assumption built in that the population from which we are
sampling, generally denoted by “N” (recall - the sample size is denoted by “n”), whether known or
not, is “large” - at least 20 times as big as the sample size - that is equivalent to saying that we are
sampling no more than 5% of the population, or, equivalently, (n/N) > .05. And, probably in 99.9%
of the cases, this is the situation.
However, there can be situations when the population size is NOT at least 20 times the sample size.
For example, say you have work at a software company that is barely out of startup phase and you have
only 20 clients. You send out a survey to all 20, but only 6 respond to the survey. Here, n = 6 and (n/N) is
.30, well above .05. We are assuming that the n = 6 is an unbiased (i.e., random) sample of the 20.
Now, there are times when this kind of low ratio of (n/N) will make a difference in both how
you calculate your conidence intervals, and the results you get. But before we show you how to do
it; we have to introduce the concept of replacement. Read on!