Flashcards in Ch22Research Deck (31):
Analysis of Variance, used to evaluate differences among two or more independent or dependent groups by partitioning the variance in the data set in different ways; also can be used to analyze differences between more than one independent variable at a time, between more than one dependent variable, or to remove the impact of an intervening variable
Two ANOVA techniques
Factorial (between subjects and mixed design) and multivariate (MANOVA and analysis of covariance ANCOVA)
Two tests to determine difference between more than one independent variable
Between-Subjects Two Way ANOVA
Mixed-Design Two Way ANOVA
Between Subjects Two Way ANOVA
also called two factor ANOVA; two IVs are examined; it can be described as a 3x2 ANOVA describing number of levels of each of the factors
Mixed Design Two Way ANOVA
used when one IV is repeated measures variable and the other IV is independent; frequently used to analyze pretest-posttest control group designs
Tests to determine differences across several dependent variables
One way ANOVA for each dependent variable, MANOVA
multivariate procedure uses an omnibus test to determine whether there are significant differences on the factor of interest when the DVs of interest are combined mathematically
Which multivariate test statistic is used most frequently?
Wilks' Lambda, converted to an estimated F statistic and the probability of is estimated F statistic is used to test the null hypothesis
How do you determine the effect of removing an intervening variable?
analysis of covariance = ANCOVA; uses overall relationship between a DV and an intervening variable, or covariate, to adjust the DV in light of the covariate scores
What are the four most common ways to analyze single-subject design data?
celeration line analysis; level, trend, slope, and variability analysis; two standard deviation band analysis; C statistic
How is a celeration line analysis used?
compares data in different phases by generating a line based on the median of subsets of data in each phase; the number of data points in the intervention phase and the number exceeding the celeration line are counted and the probability of having scores above the celeration line is generated
Level Analysis (Single-Subject Design)
celeration line is calculated; level is the difference between numerical value of observations in one phase and the numerical value of observations in a subsequent phase
Trend Analysis (Single-Subject Design)
celeration line is calculated; trend is the direction of change in the pattern of results
Slope Analysis (Single-Subject Design)
celeration line is calculated; slope is the difference between two Y values divided by the difference between the X values
Variability Analysis (Single-Subject Design)
celeration line is calculated; variability is the change in the range of scores in one phase compared with the range of scores in an adjacent phase
How is the C statistic used in single-subject designs?
appropriate for small data sets and sets with serial dependency, similar to the logic of ANOVA and the sum of squared deviation scores related to the treatment effect are divided by the sum of squared deviation scores related to the baseline
What is a survival analysis?
mathematical tool used to analyze the changing proportion of survivors over time after some naturally occurring event; there is also expanded use of survival analysis for outcomes other than death
What is the survival curve?
two defined events form the basis for a survival curve: the event that qualifies the patient for inclusion in the analysis and the event that removes the patient from the analysis (could be death but could also be a different outcome)
What is the significance of the differences between survival curves?
researchers are often more interested in the significant difference in survival curves for participants who have experienced different events or procedures versus a single event or procedure
How are confidence intervals (CIs) calculated and used?
adding and subtracting standard errors to and from the mean; the error that the researcher is willing to tolerate is based on the number of standard deviations above and below the mean that are included in the CI
When is CI testing used?
1) when sample statistics are used as estimates of population parameters 2) in addition to or instead of the results of hypothesis testing 3) means to assess the clinical importance of research results 4) with adjusted confidence levels when multiple intervals are calculated (controlling for alpha inflation) 5) when reporting the results of individual studies that are included with meta-analyses
When using the CI, when can we conclude that there is a significant difference between means?
when the CI does not contain "0"
How is the interpretation of the CI used clinically?
tells us that there is a 95% chance a population difference in means exists, this information provides sound basis on which to judge clinical importance of the finding
What does maximizing the power in a research design involve?
maximizing the size of the difference in the DV between levels of the IV, minimizing the amount of variability on the DV within levels of the IV, maximizing the sample size
How does the alpha level selected by the researcher influence power?
higher power associated with larger alpha levels
When is power analysis used?
in the design phase and after the analysis phase
How is power analysis used in the design phase?
typically 80% power used to detect differences that exist; estimate size of between-group difference that would be important, variability expect to see in groups studied, and sample size that is reasonable; dry run statistical analysis and if power <80% then reconsider elements of design
What is involved when power analysis is used to estimate sample size?
must specify their desired power level, as well as the anticipated between-group differences and within-group variability
How can effect size be used?
to help plan sample size when areas in which little previous research exists; it is the ratio of the difference between the means to the pooled standard deviation of the groups being compared; for comparison of two group means .20 is small, .50 is medium, and .80 is large
How is a power analysis used after a statistical test has failed to identify a statistically significant difference?
to look for possible explanations of non-significant findings, low probabilities of type II errors are associated with higher power values, lack of power (or a type II error) is often a likely explanation for a non-significant result in rehabilitation research