Evaluating Hypotheses
Download
Report
Transcript Evaluating Hypotheses
Testing
Hypotheses II
Lesson 10
A Directional Hypothesis (1-tailed)
Does reading to young children
increase IQ scores?
m = 100, s = 15, n = 25
sample mean also same
zobs will be the same as 2-tailed test
Differences from nondirectional
hypotheses
critical region ~
A Directional Hypothesis
1. State hypotheses
H1: m > 100
Reading to young children will
increase IQ scores.
H0: m < 100
Reading
to young children will
decrease or not change IQ scores. ~
A Directional Hypothesis
2. Set criterion for rejecting H0
a = .05, level of significance
directional (one-tailed) test
zCV = +1.645
critical
value for area = .05
in upper tail ~
Critical Regions
a = .05
zCV = + 1.645
f
-2
-1
0
+1
+2
+1.645
3. Collect sample & compute statistics
m 100, s 15
assume : X 105.5
sX
zobs
s
n
X m
sX
n = 25
15
3
25
105.5 100
5 .5
1.83
3
3
Critical Regions
a = .05
zCV = + 1.645
f
-2
-1
0
+1
+2
+1.645
4. Interpret Results
Is zobs in the critical region?
yes
reject H0, accept H1
These data suggest that reading to
young children does increase IQ.
Difference is statistically significant
but not for 2-tailed test
lower criterion than 2-tailed ~
Significance of Result
If reject H0
Statistical significance
difference between groups is ...
greater than expected by chance alone
Does NOT say it is meaningful
Even very small effects can be
statistically significant
How? ~
Significance of Result
If fail to reject H0
Data are inconclusive
Does not mean that there is no difference
Why might there be a Type II error? ~
Practical Significance
Extent to which difference is important
Magnitude of effect
Independent of statistical significance
Effect size
APA recommends it be reported
Pearson’s correlation coefficient, r
Will
cover later
Cohen’s d ~
Effect Size: Cohen’s d
Standardized measure
Units of standard deviation
General form
mean difference
d
standard deviation
For z test:
d
X m
s
Evaluating Effect Size:
Cohen’s d
For t-test:
X m
d
s
Cohen’s d
Small:
d = 0.2
Medium: d = 0.5
High:
d = 0.8
Significance Testing: Issues
Focus on H0 rather than data
H0 is always false
Small differences can be statistically
significant
Focus on results of single study
rather than accumulation
Focus on a, ignoring b
Focus on p-values misleading
Dichotomy vs continuum ~
Significance Testing: Alternatives
Criticized by some scientists
As inappropriate
Alternatives
Confidence intervals
Effect size
Meta-analysis ~