Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
113 Cards in this Set
- Front
- Back
Deductive Reasoning
|
acceptance of a general proposition, or premise, from which subswequent inferences can be drawn
|
|
Inductive Reasoning
|
invovles a type of reverse logic, developing generalizations from specific observations
|
|
Scientific Method (steps)
|
1. Unanswered question
2. Propose hypothesis 3. Conduct experiment to test hypothesis 4. Measure, analyze, interpret results 5. Conclude - support or disprove hypothesis 6. Report findings |
|
Basic (type of research)
|
fundamental, used to develop theories
|
|
Applied (type of research)
|
functional, used to dictate practice
|
|
Experimental (type of research)
|
researcher manipulates or controls select variables of interest and observes effects
|
|
Non-Experimental (type of research)
|
researcher does not manipulate or control any variables of interest
|
|
Quasi-Experimental
|
less control, no randomization and/or comparison of groups (e.g. education research)
|
|
Sequential Clinical Trials
|
continuous analysis of data as they become available (e.g. testing of new drugs)
|
|
Descriptive/Correlational
|
examination and interpretation of relationships (e.g. is caloric intake related to weight?)
|
|
Epidemiological
|
study of health determinant patterns in a population; to assess risk, prevalence of disease, disability
|
|
Evaluation
|
determination of merit or worth using criteria against a set of standards (e.g. success of program in meeting goals)
can be formative (during project) or summative (after project) |
|
Methodological
|
validation of measurement instruments or tools again accepted standards (e.g. new equipment/devices
|
|
FINER research
|
Feasibility
Interesting Novel Ethical Relevant |
|
Primary Goal of Evaluating Research Reports
|
to determine the worth or merit of a study, which ultimately depends upon the validity of the study
|
|
Key Components for an Outstanding Research Project
|
Significance
Investigators Innovation Approach Environment |
|
Inferential Statistics
|
decision making process that allows one to estimate population characteristics from sample data
|
|
Best Sample Is
|
Random
Representative Really Large |
|
Statistical Inference
|
success of this process (making valid conclusions) requires certain assumptions made concerning how well the sample represents the larger population: probability, sampling error
|
|
Sampling Error
|
tendency for sample values to differ from population values (occurs by chance)
|
|
Use of Probability in Research
|
guideline for making decisions about how well sample data estimate the characteristics of a population
are observed differences likely to be representative of population differences or occurred simply by chance |
|
Sampling
|
Define Population
Inclusion Criteria Exclusion Criteria |
|
Standard Error of the Mean (definition)
|
estimation of the population standard deviation
|
|
Standard Error of the Mean (equation)
|
Standard Error = standard deviation / the square
root of the sample size (n) |
|
Statistical Significance (definition)
|
indicates results of an analysis showing any difference or relationships are unlikely to be the result of chance
|
|
Type I Error
|
mistakenly concluding that a real difference exists, when the difference is due to chance
|
|
Type II Error
|
mistakenly concluding that a difference is due to chance when the samples represent different populations
|
|
Statistical Power (def)
|
probability that a test will detect a difference when one actually exists
probability that a test will lead to rejection of the null hypothesis |
|
Statistical Power (factors)
|
alpha level of at least p<0.05
maximize bt group differences reduce variability incr sample size use of effect size |
|
Effect Size (def)
|
used to help determine an adequate sample size (n) from pilot data, to protect against Type II errors
d = (x1-x2)/sd = effect size index |
|
Effect Size (value)
|
small effect = 0.20
medium = 0.50 (effect is half of std dev) large effect = 0.80 80% is industry standard |
|
What is an 0.80 effect size?
|
An 80% chance that we would detect a difference between the samples if one actually existed.
|
|
Confidence Interval (CI)
|
boundaries of the confidence interval are based on sample mean and its standard error
wide: greater uncertainty about the true value of population mean narrow: more certainty about the population mean |
|
Confidence Interval (calculation)
|
95% CI = mean +- (95% z-score) (standard error of mean)
|
|
Acts of Academic Dishonesty (2 Types)
|
Fabrication: creating and/or using made up/altered info in any type of scholarly activity or academic pursuit
Scientific Misconduct: deviation from the standard accepted professional practices in the performance, analysis, reporting, and/or publication of original research |
|
Human Research Guidelines (6)
|
Autonomy - pt has right to refuse or choose treatment
Beneficence - practitioner should act in best interest of pt Non-Maleficence - first, do no harm Justice - distribution of scarce health resources Dignity - pt has right to respect & ethical treatment Truthfulness/Honesty - informed consent, full disclosure |
|
Nuremberg Code:
Year Why What |
1947
Code of ethics written in response to criminal experimentation on captive victims performed during WWII Defined rules and practices for obtaining informed consent and competency of researchers |
|
Declaration of Helsinki:
Year What |
1964 (rev 1989)
Int'l Code of Ethics for Biomedical Research for development of research proposals Addressed independent review of protocol by committee not assoc with project |
|
Belmont Report:
Year What |
1979
Human protection guidelines - ensures protection of human subjects and considered std throughout the US |
|
Institutional Review Board (IRB)
Mission |
protection of human subjects and compliance with established ethical principles and regulations
|
|
IRB Review Process Considerations
|
scientific merit
feasibility of study competence of investigators subject selection, informed consent, confidentiality risk to subjects/minimized risk-benefit ratio |
|
Exempt Review
|
research deemed to pose NO RISK to subjects, generally not reviewed by IRB
collection or study of existing data, evaluation of educational programs/outcomes assessment, program evaluation, taste and food quality eval |
|
Expedited Review
|
research deemed to pose no more than minimal risk to subjects, generally reviewed by IRB subcommittee
examples: collection of non0invasice specimens, collection of data on non-invasive procedures, moderate exercise/strength test, data for group characteristics or behavior (survey) |
|
Full Reivew
|
research deemed to pose more than minimal risk to subjects, generally review by entire IRB
examples: investigational drug/clinical trials, new surgical procedures/devices, children or pregnant, individually identifiable samples |
|
Hawthorne Effect
|
observers came to check out efficiency, turns out that efficiency went up tons when the lights were on - next day, no lights, performance down - turns out that the observation itself was what was increasing performance
|
|
minimize error variance
|
standardize measurement process, equipment calibration, use only reliable and valid measurement tools/techniques
|
|
construct
|
an abstract concept used to represent unobservable behaviors or ideas by incorporating a level or scale of measurement
(i.e. intelligence, strength, pain, mood, depression, etc) |
|
nominal data
|
objects or people assigned to categories - codes have no quantitative value (i.e. 0 or 1)
(categorical) |
|
ordinal data
|
categories rank-ordered in a "greater than: less than relationship" - intervals between ranks may not be consistent and/or may not be known - only represent a position within a distribution (i.e. 1st, 2nd, 3rd)
(categorical) |
|
interval data
|
rank-order characteristics, also have known and equal distances/interval between units of measurement - relative difference and equivalence can be determined, can have negative values (i.e. temperature scales)
(continuous) |
|
ratio data
|
absolute zero point that has empirical rather than arbitrary meaning - zero represents a total absence of whatever is being measured, no negative values possible (i.e. height, weight)
(continuous) |
|
reliability
|
extent to which measurement is consistent and free from error
dependability or predictability of a specific measurement |
|
accuracy
|
nearness of a measurement to the actual value of the variable being measured
|
|
precision
|
closeness of repeated measurements of the same quantity to each other
|
|
measurement bias
|
when difference between measured value and actual value is consistently inaccurate (i.e. 5 lbs too low)
|
|
sources of error
|
instrument/equipment (test-retest)
researcher/technician (inter/intra-rater) subject/patient (intra-subject) |
|
validity
|
extent to which an instrument measures what it is intended to measure
(validity is continuum - no perfect research construct in all validity domains) |
|
repeated testing
(definition and how to control) |
prior measurement of the dependent variable may affect the results obtained from subsequent measurements
practice first, don't reveal results, randomize order |
|
content validity
|
items that make up instrument adequately sample the universe of content that defines the variable being measured
|
|
construct validity
|
ability of an instrument to actually measure an abstract concept or construct
|
|
inadequate definition of constructs
(definition and example) |
$10 word for a $1 idea - not defining the construct clearly
what is patient satisfaction? |
|
restricted generalizability
(definition and example) |
"unintended consequences" or negative consequences of the side effects of a treatment
a drug reduces pain but promotes growth of facial hair (ok for men, not women) |
|
criterion validity
|
ability of a test/instrument/measure to predict results obtained using a gold standard or criterion measure already known to be valid
|
|
descriptive stats
|
used to summarize and describe data - generalizability & external validity
|
|
measures of central tendency
|
mode, median, mean
|
|
measures of variability
|
range, variance, standard deviation
|
|
standard deviation
|
square root of the variance
|
|
1 standard deviation
|
one sided: 34%
two sided: 68% |
|
2 standard deviation
|
one sided: 47%
two sided: 95% |
|
3 standard deviation
|
one sided: 49%
two sided: 99.9% |
|
data assessment:
independent vs dependent |
independent: two or more groups consist of completely different individuals
dependent: matched pairs (same individuals tested more than once) |
|
what is the distribution that looks like a plateau? not bell curve
|
platykurtic
|
|
parametric data
|
normally distributed, analyses used are based on assumptions that population also has normal curve - increases power of analysis
|
|
nonparametric data
|
not normally distributed, analyses cannot be based on normal distributions
|
|
parametric stats - when can we use?
|
1. normal distribution
2. samples drawn at random 3. variances must be equal 4. data must be interval or ratio |
|
nonparametric stats - when do we use?
|
conditions for parametric aren't met, less powerful
|
|
data transformation
|
convert raw data by squaring, square root or calculated logarithms to make normally distributed - then can use parametric stats
|
|
concept of robustness
|
if sample is large, parametric tests can withstand SLIGHT variations from parametric assumptions
|
|
what tests are done for mean differences:
independent parametric |
independent t-test
One-Way ANOVA |
|
what tests are done for mean differences:
independent nonparametric |
Mann-Whitney U
Kruskal-Wallace |
|
what tests are done for mean differences:
dependent parametric |
paired t-test
repeated measures ANOVA |
|
what tests are done for mean differences:
dependent nonparametric |
Wilcoxon
Friedman's |
|
Why no multiple t-tests?
|
increases probability of Type I error
|
|
repeated measures
|
use one-way ANOVA to show differences among groups
|
|
post-hoc test
|
used to show which group is different
|
|
examples of post-hoc tests
|
Turkey test
Newman-Keuls test Bonferroni t-test Scheffe's comparisons |
|
Turkey test
|
Post-Hoc
|
|
Newman-Keuls test
|
Post-Hoc
|
|
Bonferroni t-test (Dunn's)
|
Post-Hoc
|
|
Sheffe's Comparisons
|
Post-Hoc
|
|
Independent t-test
|
2 groups
independent, parametric |
|
Mann-Whitney U
|
2 groups
independent, nonparametric |
|
Paired t-test
|
2 groups
dependent, parametric |
|
Wilcoxon
|
2 groups
dependent, nonparametric |
|
One-Way ANOVA
|
2+ groups
independent, parametric |
|
Kruskal-Wallace
|
2+ groups
independent, nonparametric |
|
Repeated Measures ANOVA
|
2+ groups
dependent, parametric |
|
Friedman's
|
2+ groups
dependent, nonparametric |
|
correlation
|
used to describe the relative STRENGTH of relationship between two variables
correlation coefficient (r) |
|
regression
|
used to describe predictive relationship between DV and IV
y=mb+b coefficient of determination (r2) |
|
correlation coefficient
|
goodness of fit
"r" only used for LINEAR relationships |
|
Pearson Product-Moment
|
correlation
continuous, parametric |
|
Spearman (ranks)
|
correlation
continuous/categorical, nonparametric |
|
correlation coefficient does NOT represent the _____ of association between two variables
|
percentage
|
|
one cannot infer ______ from ______
|
causation, correlation
|
|
regression
|
involves examination of two variables linearly related to be used as basis for prediction
|
|
regression involves determining a line of best fit
|
y=mx+b
|
|
what does regression line do for us?
|
help assess the accuracy of the "prediction model"
|
|
accuracy of the prediction can be determined by the ______
|
Coefficient of Determination (r2)
|
|
what is r2?
|
indicative of percentage of total variance in the DV explained by IV
|
|
if r2 value is 0.73, what does that mean?
|
73% of variance in DV can be explained by IV (or...27% of the variation involves other factors)
|
|
Standard Error of the Estimate (SEE)
|
accuracy of the predution based on variance error around regression line
farther the data points are away from line of best fit, the more error there will be, and the larger the SEE |
|
What test? Distribution - analysis of frequencies
|
Chi Square
|