Ch20Research Flashcards
What is statistics?
a discipline in which mathematics and probability are applied in ways that allow researchers to make sense of their data
What is a frequency distribution?
a tally of the number of times each score is represented in a data set
What are the four ways to represent a frequency distribution?
- frequency distribution with percentages, 2. grouped frequency distribution with percentages, 3. frequency histogram, and 4. stem-and-leaf plot
How is the central tendency represented?
mean, median, and mode
What is the variability of a data set?
the amount of spread in the data
How is variability represented?
range, variance, standard deviation
What is a normal curve?
a symmetric frequency distribution that can be defined in terms of the mean and standard deviation of a set of data
How is normal distribution represented?
z Score and percentages of the normal distribution
What is sampling distribution?
a specific type of normal distribution
How is significant difference represented?
with a null hypothesis (hypothesis that there is no statistically significant difference)
What is the Alpha Level?
Conventional level of chance tolerated is 5% (how much probability of drawing an incorrect conclusion they are willing to tolerate (pre-determined))
What information is needed for the determination of statistical probabilities?
Magnitude of the differences between groups: If within-group variability and sample size are constant, a larger between-groups difference is associated with a smaller probability that the difference occurred by chance.
Variability within a group: If the between-groups difference and sample size are constant, the differences between groups with lower within-group variability have a lower probability of occurring than the differences between group with within-group variability.
Sample size- The mean of large samples is more stable than a mean of a small sample.
What is the effect size?
the between-group difference divided by a pooled version of the SD of the groups being compared
What is the purpose for determining effect size?
allows the relative magnitude of experimental effects to be compared across variables with different measurement characteristics
What is a Type I error?
if the statistical conclusion is that there is a difference between groups when in fact there is no difference
What is a Type II error?
occurs when the statistical conclusion is that there is no difference between the groups when in reality there is a difference
What is the probability of Type II error?
beta (probability of ignoring an important difference when one exists)
What is the power of a test?
the likelihood that it will detect a difference when one exists; 1 - beta; power is high when the sampling distributions of groups have minimal overlap
What are the four major threats to statistical conclusion validity?
low power, lack of clinical importance, error rate problems, violated assumptions (and a fifth, failure to use intention-to-treat analysis)
What is a z Score?
how many standard deviations a measurement is above or below the mean
Frequency Distribution with Percentages
Absolute-number of times each score was obtained
Relative-percentage of patients who received that score
Cumulative frequency-adding relative frequencies of scores up to and including the score of interest
Revised-used when there is missing pieces of data (Adjusted or Cumulative)
Group Frequency Distribution with Percentages
Placing scores into groups (ex: 31% 33% and 35% would all be placed into 30-39% group)
Disadvantage: information is lost
Frequency histogram
height of bar represents frequency of observations in group
Disadvantage: information is lost
Stem and leaf plot
presents data concisely without losing information in the grouping process
What is central tendency
A set of data collapsed into a single score that represents the whole set, “descriptive statistics”
What is inferential statistics?
Used to determine whether or not there are significant differences between groups
Alpha Level of Inflation?
When researchers contact many tests within a given study, which increases the overall probability of obtaining chance significant differences. Can be corrected by using a more stringent alpha level for each individual test.
Four ways to increase power
- Maximize between-groups differences by controlling extraneous variables and making sure they apply experimental techniques constantly (rigid controls may reduce a study’s external validity because of how much the setting is being controlled)
- Reduce within-group variability by studying homogenous groups of subjects or by using subjects as their own controls in repeated measurement designs (these strategies may also reduce external validity by narrowing the group of patients to whom the results can be generalized
- Increase sample size.
- Select a more liberal alpha level at which to conduct their statistical tests (this method is unlikely to be acceptable in the scientific research community judging the research.
What are power tables used for?
To determine what sample size is needed to achieve a particular level of power.
Variables needed: desired power, alpha level used, estimate of between-groups difference that is considered clinically meaningful and an estimate of the within-group variability expected
