more on inferential statistics - koç university stats.pdf · inferential statistics copyright ©...
TRANSCRIPT
3/16/14
1
More on Inferential Statistics
Inferential Statistics
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 2
• What is Significant? • An inferential statistical test can tell us whether the results of an
experiment can occur frequently or rarely by chance. • Inferential statistics with small values occur frequently by chance. • Inferential statistics with large values occur rarely by chance.
• Null Hypothesis • A hypothesis that says that all differences between groups are due to
chance (i.e., not the operation of the IV). • If a result occurs often by chance, we say that it is not significant and
conclude that our IV did not affect the DV.
• If the result of our inferential statistical test occurs rarely by chance (i.e., it is significant), then we conclude that some factor other than chance is operative.
t-test
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 3
• t Test • The t test is an inferential statistical test used to evaluate the
difference between the means of two groups.
• Degrees of Freedom • The ability of a number in a specified set to assume any
value.
3/16/14
2
The t Test
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All
rights reserved. 4
• Interpretation of t value • Determine the degrees of freedom (df) involved.
• Use the degrees of freedom to enter a t table. • This table contains t values that occur by chance.
• Compare your t value to these chance values.
• To be significant, the calculated t must be equal to or larger than the one in the table.
One-Tail Versus Two-Tail Tests of Significance
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 5
• Directional versus Nondirectional Hypotheses • A directional hypothesis specifies exactly how (i.e., the direction)
the results will turn out.
• A nondirectional hypothesis does not specify exactly how the results will turn out.
One-Tail Versus Two-Tail Tests of Significance
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 6
• One-tail t test
• Evaluates the probability of only one type of outcome (based on directional hypothesis).
• Two-tail t test
• Evaluates the probability of both possible outcomes (based on nondirectional hypothesis).
3/16/14
3
The Logic of Significance Testing
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 7
• Typically our ultimate interest is not in the samples we have tested in an experiment but in what these samples tell us about the population from which they were drawn.
• In short, we want to generalize, or infer, from our samples to the larger population.
The Logic of Significance Testing
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 8
A. Random samples are drawn from a population.
B. The administration of the IV causes the samples to differ significantly.
C. The experimenter generalizes the results of the experiment to the general population.
Type I and Type II Errors
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 9
• Type I (alpha) Error – false positive • Accepting the experimental hypothesis when the null hypothesis
is true.
• The experimenter directly controls the probability of making a Type I error by setting the significance level.
• You are less likely to make a Type I error with a significance level of .01 than with a significance level of .05
• However, the more extreme or critical you make the significance level to avoid a Type I error, the more likely you are to make a Type II (beta) error.
3/16/14
4
Type I and Type II Errors
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 10
• Type II (beta) Error – false negative • A Type II error involves rejecting a true experimental hypothesis.
• Type II errors are not under the direct control of the experimenter.
• We can indirectly cut down on Type II errors by implementing techniques that will cause our groups to differ as much as possible. • For example, the use of a strong IV and larger groups of
participants.
Type I and Type II Errors
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 11
Example
• Suppose a researcher comes up with a new drug that in fact cures AIDS. She assigns some AIDS patients to receive a placebo and others to receive the new drug. The null hypothesis is that the drug will have no effect.
• In this case, what would be the Type I error?
• In this case, what would be the Type II error?
• Which do you think is the more costly error? Why?
3/16/14
5
TYPE I AND TYPE II ERRORS
• The probability of making a Type I error – Set by researcher
• e.g., .01 = 1% chance of rejecting null when it is true
• e.g., .05 = 5% chance of rejecting null when it is true
– Not the probability of making one or more Type I errors on multiple tests of null!
• The probability of making a Type II error – Not directly controlled
by researcher – Reduced by increasing
sample size
MAKING A DECISION
If You… When the Null Hypothesis Is Actually… Then You Have…
Reject the null hypothesis True (there really are no differences)
Made a Type I Error
Reject the null hypothesis False (there really are differences)
Made a Correct Decision
Accept the null hypothesis False (there really are differences)
Made a Type II Error
Accept the null hypothesis True (there really are no differences)
Made a Correct Decision
Example
• In the criminal justice system, a defendant is innocent until proven guilty. Therefore, the null hypothesis is: “This person is innocent.” • In this case, what would be the Type I error?
• In this case, what would be the Type II error?
• Which do you think is the more costly error? Why?
3/16/14
6
Effect Size
Copyright © 2013 Pearson Education, Inc., Upper Saddle River, NJ 07458. All rights reserved. 16
• Effect Size (Cohen’s d ) • The magnitude or size of the experimental treatment.
• A significant statistical test tells us only that the IV had an effect; it does not tell us about the size of the significant effect.
• Cohen (1977) indicated that d values greater than .80 reflect large effect sizes.