Transcript Choosing an Appropriate Statistical Test Procedure

LIS 397.1
Introduction to Research in
Library and Information
Science
Choosing an Appropriate
Statistical Test Procedure
R. E. Wyllys
Copyright 2003 by R. E. Wyllys
Last revised 2003 Jan 15
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science
How to Choose a Standardized
Statistical Test
• Choice depends on your real-life
problem
• Figure out how many variables are
involved
– If just one variable, decide how many
populations (groups) are involved
– If two variables, consider nature of
variables: interval or categorical
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science
How to Choose a
Standardized Statistical Test
• 1-Variable Problems: Is your question:
– Does the variable have a particular mean value in
some population? Use
• Confidence interval
– Does the variable average the same (i.e., does it
have the same mean) in two populations? Use
either
• t-Test for dependent samples (if pairs of observations can
be made on each element of your sample); else, t-Test
for independent samples; or
• ANOVA with repeated values (if paired observations can
be made); else, ANOVA without repeated values
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science
How to Choose a
Standardized Statistical Test
• 2-Variable Problems: Is your question:
– Is there some correlation or association between
the variables? Use
• Pearson correlation if variables are interval type
• Chi-square test of association if variables are categorical
type
– Is it possible to estimate (predict) the value of one
of a pair of interval variables on the basis of
observing the value of the other? Use
• Linear regression
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science
After You Choose a
Standardized Statistical Test
• Choose the level of significance (or the confidence
level) at which you want to work
• Make your observations and enter them into an
appropriate computer program
• Using the program, apply your chosen test to your data
• Note the resulting values of your test statistic and the
probability or significance that the program reports for
the test statistic
• Accept or reject the null hypothesis that is built into the
procedure you have used.
• Interpret your decision with respect to your real-world
problem
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science
Statistics and Statistical Tests Can Be
Valuable Tools in Solving Problems
School of Information - The University of Texas at Austin
LIS 397.1, Introduction to Research in Library and Information Science