Study Guides
(238,467)

Canada
(115,151)

University of Toronto St. George
(7,977)

CSB345H1
(9)

William Navarre
(9)

# Session 5 - What does 'not significant' really mean.docx

Unlock Document

University of Toronto St. George

Cell and Systems Biology

CSB345H1

William Navarre

Fall

Description

HMB325H © Lisa| Page 113
S E S S I O N 5 : W H AT D O E S ' N O T S I G N I F I C A N T ' R E A L LY M E A N ?
( C H A P T E R 6 )
LEARNING OBJECTIVES
1. recognize the relevant study question
2. recognize & describe the characteristics of the data (i.e. types of variables & levels of
measurement) associated w the study
3. describe the descriptive statistics to be used
4. identify & carry out the appropriate statistical test & determine the associated probability (P)
5. discuss the factors influencing study power & sample size & how they interact
6. describe how the width of a confidence interval are influenced by a study’s power
7. calculate study power for t-tests & difference of proportion analyses
THE ROLE OF CHANCE
• studies are done to advance & inform biomedicine/healthcare
• relevant data needs to be collected in these studies
• statistical tools enable 2 study tasks:
summarize the data (descriptive statistics)
measure the role of chance (statistical testing)
• the role of chance?
is an observed study result real?
does it indicate a real effect or real difference
this can’t be determined directly, so we examine other possible explanations like chance
to see what role they might play
what is the probability of observing study results as extreme as those by chance?
is random variation/chance a reasonable explanation for the observed results?
need to identify & use the applicable statistical test, which returns the probability (P) of
study results as extreme as these arising by chance
WHICH STATISTICAL TEST?
1. level of measurement of outcome variable: continuous, dichotomous, discrete,
nominal, ordinal
2. type of comparison/study design: 2 groups OR 2+ (independent) groups compared,…
Type of Comparis
Test data on
ANOVA quantitative 2+ groups
*
quantitative
t test * 2 groups
z test dichotomou 2 groups
s
χ test nominal ≠ 2+ groups
Fisher’s exact dichotomou
# 2 groups
test s
≠discrete or continuous, & normally distributed
includes dichotomous
#if expected value of any cell is <5
• all statistical tests answer the question: what is the probability of results as extreme as
these occurring simply by chance?
• all tests answer this question by examining the ratio of the observed results to those that
would be expected simply by chance:
observedresults
ratio=
expectedresults HMB325H © Lisa Z| Page 2
• this ‘ratio’ is called the test statistic
• the larger it is, the less probable that the study’s results might have arisen by chance
• to determine the applicable probability (P) value, find the ratio value in the applicable
statistical table
• this probability never goes to 0
every result has some probability of arising by chance
• how probably do study results need to be for us to conclude they are due to chance?
• how improbably do study results need to be for us to conclude that chance is not a
probably explanation?
HOW IMPROBABLE?
• how low (improbable) should the probability (P) be for use to conclude that the study
results are indicative of a ‘real’ effect?
this is the alpha (α) level selected for the study
• α level is balanced bw 2 unwanted outcomes:
1. false-positive study (type I error): wrongly concluding there is an effect
2. false-negative study: wrongly concluding there isn’t an effect
• possible study outcomes:
the truth
+ –
false-
true-
+ positive positive
study (type I error)
result false-
– negative true-
negative
(type II error)
• 2 explanations for ‘not significant’:
1. there truly is no effect (though we can’t determine this directly)
2. there is a true effect, but the study didn’t

More
Less
Related notes for CSB345H1