correct (American College of Physicians—American Society of Internal Medi-
cine, 2001; Hoekstra, Johnson, & Kiers, 2012). In other words, CIs estimate the
degree of confidence one can have about the inferences. Researchers typically
report confidence levels of 95% or 99%.
Use of inferential statistics to test hypotheses is best suited to research ques-
tions or hypotheses that ask one of two broad questions:
» Is there a difference between the groups?
» Is there a relationship among the variables?
Most experimental and quasi-experimental designs involve questions asking
if there is a difference between the groups. For example, the use of inferential
statistics to test the hypothesis “Patients who have uninterrupted sleep cycles
have better wound healing than do patients who awaken throughout the
night” would be appropriate. Research questions or hypotheses that ask about
relationships among variables, such as “Is there a relationship between the
number of hours of sleep and a score on a memory exam?” can also be tested
using inferential statistics.
When deciding which statistical tests to use to analyze data, researchers must
take into account many factors (Hayes, 1994). After considering whether the
research questions or hypotheses involve groups or variables, the next most
important factor researchers consider is the level of measurement. Whether
variables are nominal, ordinal, interval, or ratio is important because some tests
are appropriate for interval and ratio data but not for other levels of measure-
ment. Other factors that can influence selection of inferential statistical tests
include whether: (1) a probability sampling method was used, (2) the data are
normally distributed, and (3) there is potential confounding of the variables.
Nurses should keep in mind that the strongest inferences can be made when
the level of measurement is interval or ratio, a probability sampling method
was used, the sample size is adequate, and the data are normally distributed.
It’s All About Chance
Regardless of the type of question being asked, the major unanswered question
is: What is the likelihood that the findings could have occurred by chance alone?
For example, suppose that you toss a coin 10 times, and each time it lands heads
up. Could that happen by chance? Absolutely. Now suppose you flip it another
10 times and it lands with heads up every time. Although it is possible that a
coin could land heads up 20 times in a row, this would
be a rare occurrence. Would you begin to wonder if this
were happening by chance or would you suspect that the
coin is not fair? How many times would you want to toss
the coin before you conclude that the coin is not fair?
FYI
Statistics enable nursing researchers to de-
termine the probability that results are not
a result of chance alone.13.6 Inferential Statistics: Can the Findings Be Applied to the Population? 353