Power Analysis

POWER ANALYSIS

DEFINITION
 Power analysis is the procedure that researchers can use to determine if the test contains enough
power to make a reasonable conclusion.
 From another perspective power analysis can also be used to calculate the number of samples required
to achieve a specified level of power.
 A power analysis is the calculation used to estimate the smallest sample size needed for an experiment.
 It helps to determine if a result from an experiment or survey is due to chance, or if it is genuine and
significant.
PARTS OF STATISTICAL POWER
Statistical power is made of four related parts.
1. Effect size: The quantified size of a result present in the population.
2. Sample size: A number of things measured: How many dogs do we need to observe?
3. Significance: The level of significance used in the experiment, generally 5 percent.
4. Statistical power: The probability of accepting the alternative hypothesis.

TYPES OF STATISTICAL POWER ANALYSIS

 Power analysis typically takes two forms, priori power analysis and post-hoc power analysis.
 In priori power analysis, power is calculated before conducting significance tests evaluating the hypothesis
interest, whereas in post-hoc power analysis, power is calculated after conducting these significance tests.
1. Priori power analysis
 Priori power analysis considered the ideal type of power analysis by most authors.
 In a priori power analysis, researchers specify the size of effect to be detected (i.e., measure of the “distance”
between H0 and H1), the “α” level and the desired power (1-β) of the test.
 Given these specifications it is possible to compute the necessary sample size “n”.
 For the purposes of designing a research study, priori power analysis is of great utility.
2. Post-hoc power analysis
 Power analyses can be useful during the design stage of a study.

 But sometimes power analyses are run after the data have been analysed and especially when insignificance
results occurred.
 Unlike a priori power analysis that relies on estimates of effect size, post-hoc power analysis conveys the actual
power in the study through the observed effect size rather than an estimated value.
THE PROBLEMS CAUSED BY INSUFFICIENT POWER
 Power should be a primary consideration in many decisions in the design and implementation of a statistical
study. Lack of a power may lead to erroneous decisions concerning the null hypothesis.
 A deficiency in power increases the probability that a type II error will occur.
 In other words, low power increases the likelihood that a researcher will make a decision to retain a false null
hypothesis
 Studies that conducted with sufficient statistical power provide to be successful enough chance to researchers.
THE PROBLEMS CAUSED BY OVERPOWERED STUDIES

 Some authors have issued a warning against the potential for "overpowered" research, which could waste
funding resources or result in the likelihood of identifying insignificant effects.
 However, in some circumstances, a little extra power could be helpful. For primary study outcomes, funding
organisations frequently want a power of 0.80 at a least.
G-POWER: A GENERAL POWER ANALYSIS PROGRAM
 In the past, researchers have used imprecise methods to decide required sample sizes rather than using priori
power analysis.
 These methods have many times shaped on budget of research, time limitation and recommendations of experts.
Cohen (1992) said that “researcher may find power analysis too complicated and may simply avoid the issue.”
 For that reason, power tables were created by Cohen in order to facilitate calculation of statistical power.
 In addition, computer software programs were created by many experts. G-Power is one of them.
 G-Power is a free software program to assist researchers in conducting power analyses either priori or post-hoc.
CALCULATION OF POWER
To estimate statistical power, you need the following information:
 The statistical test to be used (e.g., independent samples t test, correlation, paired samples t test,
etc.)
 The (estimated) population effect size (often given in terms such as Cohen’s d, eta squared, f, and
many other effect size indices). A reasonable guess can sometimes be made by looking at reported
effect sizes for studies similar to yours.
 The planned alpha level (and whether the test is one-tailed or two tailed, where applicable).
 For some analyses, such as ANOVA, the number of groups to be compared (or the df for the
between group effect).
Using software
Power analysis is an approach to determine the minimum sample size for quantitative research - including
participant selection for data collection. Power analysis is often conducted using software, such as
G*Power3, and incorporates multiple assumptions/factors, including:
 Alpha
 Effect
 Power
 Statistical test
 Number of groups
Once determined, then the number of invitations to participate can be determined, based on the assumed
response rate. For instance, if the minimum sample size determined via power analysis is 100, and a 10%
survey response rate is assumed (from prior studies), then a population sample of 1000 would be needed.
Tables
When you look up the table entry for statistical power in the appropriate statistical power table, and look at
the line for your sample size, you find an estimated statistical power. A conventional standard for adequate
power is .80, but this is arbitrary.

Power Analysis

Uploaded by

Copyright:

Available Formats

Power Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Power Analysis

Uploaded by

Copyright:

Available Formats

POWER ANALYSIS

TYPES OF STATISTICAL POWER ANALYSIS

1. Priori power analysis

2. Post-hoc power analysis

 Power analyses can be useful during the design stage of a study.

THE PROBLEMS CAUSED BY INSUFFICIENT POWER

THE PROBLEMS CAUSED BY OVERPOWERED STUDIES

G-POWER: A GENERAL POWER ANALYSIS PROGRAM

To estimate statistical power, you need the following information:

You might also like