Microsoft Office Tutorials and References

In Depth Information

6.8 Summary

The use of inferential statistics is invaluable in analysis and research. Inferential

statistics allows us to infer characteristics for a population from the data obtained

in a sample. We are often forced to collect sample data because the cost and

time required in measuring the characteristics of a population can be prohibitive.

In addition, inferential statistics provides techniques for quantifying the inherent

uncertainty associated with using samples to specify population characteristics. It

does not eliminate the uncertainty due to sampling, but it can provide a quan-

titative measure for the uncertainty we face about our conclusions for the data

analysis.

Throughout Chap. 6 we have focused on analyses that involve a variety of data

types—categorical, ordinal, interval, and rational. Statistical studies usually involve

a rich variety of data types that must be considered simultaneously to answer our

questions or to investigate our beliefs. To this end, statisticians have developed a

highly structured process of analysis known as tests of hypothesis to formally test

the veracity of a researcher’s beliefs about behavior. A hypothesis and its alterna-

tive are posited and then tested by examining data collected in observational or

experimental studies. We then construct a test to determine if we can reject the null

hypothesis based on the results of the analysis.

Much of this chapter focused on the selection of appropriate analyses to per-

form the tests of hypothesis. We began with the chi-squared test of independence

of variables. This is a relatively simple, but useful, test performed on categor-

ical variables. The z-Test and t-Test expanded our view of data from strictly

categorical, to combinations of categorical and interval data types. Depending on

our knowledge of the populations we are investigating, we execute the appro-

priate test of hypothesis, just as we did in the chi-squared. The t-Test was then

extended to consider more complex situations through ANOVA. Analysis of vari-

ance is a powerful family of techniques for focusing on the effect of independent

variables on some response variable. Finally, we discussed how design of experi-

ments helps reduce ambiguity and confusion in ANOVA by focusing our analyses.

A thoughtful design of experiments can provide an investigator with the tools

for sharply focusing a study, so that the potential of confounding effects can be

reduced.

Although application of these statistics appears to be difﬁcult, it is actually very

straight forward. Table 6.16 below provides a summary of the various tests presented

in this chapter and the rules for rejection of the null hypothesis.

In the next chapter we will begin our discussion of
Model Building
and

Simulation
—these models represent analogs of realistic situations and problems that

we face daily. Our focus will be on
what-if
models. These models will allow us to

incorporate the complex uncertainty related to important business factors, events,

and outcomes. They will form the basis for rigorous experimentation. Rather than

strictly gather empirical data, as we did in this chapter, we will collect data from our

models that we can submit to statistical analysis. Yet, the analyses will be similar to

the analyses we have performed in this chapter.

Search JabSto ::

Custom Search