Flashcards in Ch21Research Deck (36):
What are the three most common distributions used for statistical tests?
t, F, and chi-squared distributions
What are parametric tests?
based on specific assumptions about the distribution of populations; use sample statistics such as the mean, standard deviation, and variance to estimate differences between population parameters
What are nonparametric tests?
use rank or frequency information to draw conclusions about differences between populations, not based on specific assumptions about the distributions of populations (parametric are often more powerful and preferred)
What are the assumptions of parametric testing?
random selection, homogeneity of variance, measurement level of the data as interval or ratio
When are data sets independent?
when values in one set tell nothing about values in another set
When are data sets dependent?
when the sets of numbers consist of repeated measures on the same individuals, can be different individuals when they are matched for other factors
What are the 10 steps in the statistical testing of differences?
1) state the null & alternative hypotheses in parametric terms 2) decide alpha level 3) determine independent or dependent samples 4) determine if parametric assumptions met 5) determine appropriate statistical test 6) calculate the statistic 7) determine degrees of freedom 8) determine probability of obtaining test statistic given the df 9) compare probability of obtaining test statistic with alpha 10) evaluate statistical conclusion in light of clinical knowledge
What is assumed in the z distribution?
population standard deviation is known
Why can z distributions not normally be used?
the population standard deviation is not normally known so researchers use distributions that resemble the normal distribution but altered to account for errors that are made when population parameters are estimated (t, F, and chi-square distributions)
What distinguishes a t distribution from a z distribution?
greater proportion in the tails and lesser proportion in the center, spread to account for errors when population is estimated from sample statistics, becomes more similar to z distribution with larger sample sizes and increased degrees of freedom
What is used to calculate degrees of freedom?
number of participants in study or number of levels of independent variable or both
What distinguishes the F distribution?
squared t statistics, asymmetric and only positive values, actual shape depends on two different df: one with number of groups being compared and one associated with sample size
What distinguishes the Chi-Square distribution?
squared z scores, shape varies with degrees of freedom
What are two major classes of parametric tests?
t tests and ANOVAs
What are two strategies for non-normal distributions?
convert the data mathematically to become more normally distributed, or use non-parametric test which does not require normal distribution
What is the best strategy to ensure homogeneity of variance?
have equal or nearly equal sample sizes across groups (differences in variance become less of a concern when sample sizes of groups being compared are the same)
Which type of data would be appropriate for non-parametric tests?
nominal and ranked ordinal data (interval and ratio data can be converted into ranks or grouped into categories as well)
Which type of data is appropriate for parametric tests?
interval and ratio (need means and variances)
Can parametric tests be used for ordinal data?
yes, as long as the distribution of ordinal data is approximately normal and parametric assumptions are met
What tests are used for differences between two independent groups?
independent t test, Mann-Whitney or Wilcoxon rank sum test, chi-square test of association
What is the independent t test?
ratio of differences between groups to the differences within groups, test statistic formula partitions the variability in data into explained and unexplained variability (unexplained is the variance within groups or the denominator)
When variability explained by the IV is large, will the test statistic of the independent t test be large or small?
What must be established about the hypothesis before conducting the independent t test?
differentiation between directional and non-directional hypothesis, directional is only justified if there is existing evidence of the direction of the effect
What are the Mann-Whitney and Wilcoxon rank sum tests?
non-parametric alternatives to the independent t test, scores are ranked from the two groups regardless of group membership, used if assumptions of independent t test are violated
What is the chi-square test of association?
concerned with proportion of patients with a particular score, convert raw score into categories i.e. "less than" or "greater than or equal to," used to determine whether two groups had similar proportions of patients in categories
What tests are used to determine differences between two or more independent groups?
one-way ANOVA, Kruskal-Wallis test, Chi-Square test of association
What distinguishes a one-way ANOVA?
only one IV examined, partitions variability in a sample into between groups and within groups variability, F statistic generated (squared t), if no significant difference then test is done
If omnibus F test tells us there is a significant difference between means, then what must be done next?
conduct a multiple-comparison test called a contrast, ex. include planned orthogonal contrasts, Newman-Keuls test, Tukey test, Bonferroni test, Scheffe test
What is the Kruskal-Wallis test?
non-parametric equivalent of the one-way ANOVA, ranks the scores regardless of group membership, distribution of KW statistic approximates a chi-square distribution, use Mann-Whitney test (multiple comparison procedure) with a Bonferroni adjustment of alpha if KW test is significant
What tests are used to determine the differences between two dependent samples?
paired t-tests, Wilcoxon signed rank test, McNemar test
What is the paired t-test?
used to determine the difference between each pair of measurements, mean difference and SD of differences calculated, determine probability of given t statistic
What is the Wilcoxon signed rank test?
non-parametric version of the paired t-test, if no difference then sum of positive and negative ranks should be approximately equal to sum of negative ranks
What is the McNemar test?
nominal data analogue to paired t-test and Wilcoxon signed rank test, dependent samples version of the chi square test, can only be used to analyze 2x2 contingency tables so usefulness is limited
What tests are used to determine differences between two or more dependent samples?
repeated measures ANOVA or Friedman's ANOVA
What is the repeated measures ANOVA?
extension of paired t-test to more than two dependent samples, between subject and within subject categories then generate F ratio, post hoc tests to determine which group is significantly different, if significance is found then some suggest using paired t-test with Bonferroni adjustment of alpha