Many researchers state that statist...

Many researchers state that statistical power is an important general [i]or[/i] abstract notion that should be considered during the design and analysis phases of an experiment. Institutional review boards may entreat that investigators address the issue of statistical power when an application is submitted for review. In addition, it is an increasingly often met with practice for article reviewers to entreat analysis clarifying the statistical power of a cogitation being considered for journal publication. Nevertheless, analysis of studies published in rehabilitation,' nursing and physiology' journals indicate a widespread question at issue with inadequate power. It has become necessary to have a clear understanding of statistical power in order to critically review the literature.

The meaning of this article is to provide an overview of the conception of statistical power, the factors affecting statistical power, and suggestions for interpreting be deriveds from a statistical power viewpoint. For a comprehensive treatment of the control it is recommended that the reader confer a text devoted to the subject"' Several research body s are recommended for a brief review of statistical power/''

EXPERIMENTAL STUDIES

Experimental research uses studies to investigate the probability of cause-effect relationships. Since it is rarely possible to example every individual in the population of interest, a sample or subset of individuals is usually studied. The data generated from the sample is an estimate of the data of the entire population. In addition, any measurements involve more [i]or[/i] less degree of error and are, therefore, solitary estimates of the true values." Since there is always an doubt about whether sample data generated during an experiment contemplates the true events, statistical proofs are used to generate a likelihood or probability that a conclusion might be faithful Most statistical tests evaluate the difference between cluster (eg, treatment versus control) result means to the variability in issues within groups in order to determine p values which indicate statistical significance.

In order to determine cause-effect, common independent variable is manipulated by way of the experimentor and its weight on the outcome (dependent) variable is measured. For example, individual might ask if inspiratory muscle training (IMT) improves exercise tolerance in individuals with spinal cord injury (SCI). The independent variable could be called 'Intervention' and individuals would either receive the intervention (experimental group) or not (control group) Since there are 2 options available, the 'Intervention' variable is said to have 2 plains The dependent variable, exercise tolerance, could be measured by way of having subjects perform a wheelchair ergometer example After administering the inspiratory muscle training for a specified time, the data for the pair groups is statistically analyzed and a p value is generated. Usually, if the p value is

The statements in the above paragraph might lead single in kind to believe that interpretation of issues is straightforward. However, several questions may arise. in what manner much of a difference in exercise tolerance is necessary between enslaves in the control and experimental form into groupss in order to obtain statistical significance (p

DEFINING STATISTICAL POWER

Statistical power is defined as the probability that a statistical analysis will lay open a difference which actually exists' or in other words, the alternative (research) hypothesis is accepted and it is actually true" Conceptually, power is 1-[beta], where [beta] is the likelihood of committing a sign Il error. A Type Il error is acceptance of the characterless hypothesis when, in fact, a difference actually exists/ symbol I and Il errors are inversely related to each other. Decreasing the significance criterion, [alpha], from 005 to 001 will increase the probability of causing a protoplast Il error and, therefore, decrease power. However, exemplar Il errors are considered to be les serious than sign I errors. For example, the 'innocent before proven guilty' assumption underlying the U legal a whole is consistent with the idea that it is preferable to acquit someone who is actually guilty (a archetype Il error) than to convict someone who is actually innocent (a shadow I error). Use of [beta] values between 005 and 020 are usually praiseed which indicates a 5% to 20% likelihood of committing a shadow Il error. Therefore, the praiseed minimum statistical power is 080''' In other words, there is an 80% probability that the consideration will yield a statistically significant difference which with truth exists.

FACTORS INFLUENCING POWER

The factors which influence power are the consequence size (the difference in issues between groups), sample variance (wimin assign places to variability in the outcome measure), sample size, and the criterion for significance (a).7 8 Tables which relate these factors can be used to determine common unknown factor.4 For example, when planning a consideration one can determine how many controls are required in order to lay open an estimated treatment effect, at p