Exam III Flashcards
What are the two key attributes of data measurement (variable)?
- magnitude
2. consistency of scale/ fixed interval
What is the third variable used to assess data when the first two are answered with yes’?
-rational/absolute zero
What are the three levels of measurement?
- nominal
- ordinal
- interval
Examples of nominal measurements
-variables that are simply labeled variables without quantitative characteristics
- gender
- hair color
- occupation
examples of ordinal variables
-variables that have magnitude but no consistency of scale
- interval of ages (1-18, 19-50)
- months homeless (less than three, greater than three)
examples of interval data
-number with units at the end
- age
- number of siblings
What variables are considered discrete vs continuous?
discrete: nominal, ordinal
continuous: interval
What are the measures of central tendency utilized for describing continuous data?
- mode/mean/median
- outliers
- min/max/range
- interquartile range
What is variance?
difference in each individual measurement value and the groups’ mean
What is standard deviation?
- square root of variance value
- know eqn
How do you know when a graph is normally distributed?
-the mean/median/mode are near equal
What are parametric tests?
-stats tests useful for normally distributed data
What are the two types of graphical shapes?
- positively skewed
- negatively skewed
How can you tell the difference between a positively and negatively skewed graph based on stats alone?
positively skewed:
- mean is greater than median
negatively skewed:
-mean is less than median
Definition of skewness
-a measure of asymmetry of a distribution
+a perfectly normal distribution is symmetric and has a skewness value of 0
What is kurtosis?
- a measure of the extent to which observations cluster around the mean. For a normal distribution, the value of the kurtosis statistic is 0.
- how peaked the graph is
positive vs negative kurtosis
positive -> more cluster
negative -> less cluster
What do you get when you add or subtract a std dev from the mean?
-range of middle 68%, 95%, and 99%
What are the required assumptions for interval data to select a parametric test?
- normally distributed
- equal variances
+Levene’s test - randomly derived and independent
What is the Levene’s test?
-a test for variablity
How does one handle interval data that is not normally distributed?
- use a statistical test that does not require the data to be normally distributed (non-parametric)
- transform data to a standardized value (z-score or log)
What is power?
1-beta (type 2 error)
- the ability of a study design, its methodology, and the selected test statistic to detect a true difference if one truly exists between group comparisons (analogous to sensitivity)
- researchers typically choose 80% power to truly distinguish a difference between two groups
What does power have to do with sample size?
-the larger the sample size, the greater the likelihood of detecting a difference if one truly exists (increase in power)
What needs to be determined with each sample? (for size)
- minimum difference between groups deemed significant
+the smaller the difference between groups necessary to be considered significant, the greater the N needed - expected variation of measurement
- alpha and beta error rates
+alpha -> type 1 (5%)
+beta -> type 2 (20%)
+add in anticipated drop outs or loss to follow ups
What is the p value and why is it important?
-statistical tests that determine the possible differences or relationships between variables
-same as type 1 error and alpha
-if less than 0.05, then you can claim there is a difference between groups
+small chance that the difference didn’t occur by chance
When the p value is high which hypothesis will you be more likely to accept?
null hypothesis
What is type 1 error (alpha)?
- false positive
- rejecting the null hypothesis when it is actually true
What is type 2 error (beta)?
- false negative
- accepting the null hypothesis when you should have rejected it
When comparing baseline characteristics in a study what you want the p values to be?
- high p values, want no difference between groups
- same for Levene’s test for baseline
How would one interpret the p value in words?
-probability of making a type 1 error if null is rejected
What is the confidence interval?
-percentage of confidence that statistically the real difference or relationship resides
-based on:
+variation in sample (v/sd)
+sample size
What is the most commonly selected percentage for a CI?
95%
How is a CI interpreted?
-we are 95% confident that the true difference/relationship between the groups is contained within the confidence interval range
How are CI interpreted without the p value?
- if CI crosses 1.0 (OR/RR/HR) or 0.0 (other), then the groups are not significant
- same as if p were greater than 0.05
CI on a Forest plot
-the error bars represent the CI
What are the 4 key questions for selecting the correct statistical test? Which two are more for comparing frequencies?
- What type of data is being collected? (nominal, ordinal, interval)
- What type of comparison/assessment is desired?
- How many groups are being compared?
- Is the data independent or related(paired)?
-3,4 deal with frequency comparisons
What are the different types of comparisons/assessments?
- correlations
- event-occurrence/time to event -> survival test
- outcome prediction/association (OR) -> regression
What is a correlation?
-provides a quantitative measure of the strength and direction of a relationship between variables
-values range from -1 to 1
+perfectly negative = -1
+perfectly positive = +1
What is a partial correlation?
-a correlation that controls for confounding variables
What are the difference correlation tests that can be run for each individual data type?
nominal -> contingency coefficient
ordinal -> spearman correlation
interval -> pearson correlation
What does a greater than 0.05 for a Pearson correlation mean?
-there is no linear correlation, there can still be non-linear correlations present
What are survival tests?
-compares the proportion of, or time to, event occurrences between groups
What are the different survival tests for the types of data?
nominal -> log-rank test
ordinal -> cox-proportional hazards test
interval -> Kaplan-Meier test
*all can be represented by a Kaplan-Meier curve
What is a regression? What can be calculated from this type of test?
- provide a measure of the relationship between variables by allowing the prediction about the dependent, or outcome, variable knowing the value/category of independent variables
- OR can be calculated from this for a measure association
What are the different types of regression tests for each type of data?
nominal -> logistic regression
ordinal -> multinominal logistic regression
interval -> linear regression
For nominal data what tests can be run for 2 groups of independent data?
(Pearson’s) chi squared
For nominal data, what test can be run for more than three groups of independent data?
chi squared test of independence
For nominal data, what test can be run for more than or equal to two groups with expected cell count of less than five?
-Fisher’s Exact test
What are the assumptions for a chi squared test?
- usual chi square distribution for nominal type data
- no cell with expected count of less than 5
What are the two hypotheses one makes before starting a study?
Null hypothesis: states that there will be no true difference between groups compared
Alternate hypothesis: there will be a difference between groups
After a chi squared test is conducted in three independent groups to determine a difference between the groups. What test do you use to figure out where the difference lies?
- post hoc test
- multiple chi squared tests introduce increase in type 1 error
What is the Bonferroni test of inequality?
-adjusts the p value for # of comparisons being made, conservative
Test for nominal 2 groups of paired/related data?
McNemar test
Test for nominal 3 or more of paired/related data?
-Cochran test
What are the key words for paired/related data?
- pre vs post
- before vs after
- baseline vs end
Test for ordinal 2 groups of independent data?
Mann-Whitney test
Test for ordinal three or more groups of independent data?
-Kruskal-Wallis test
+compares the median values between groups
+use post hoc if group comparison is significant
Test for ordinal 2 groups of paired/related data?
-Wilcoxon Signed Rank test
Test for ordinal 3 or more paired/related data?
-Friedman test
What are the three post hoc tests for ordinal data? What are the differences?
-Student-Newman-Keul test
+compares all pairwise comparisons possible
+all groups must be equal in size
-Dunnett test
+compares all pairwise comparisons against a single control
+all groups must be equal in size
-Dunn test
+compares all pairwise comparisons possible
+useful when all groups are NOT of equal size
Test for interval 2 groups of independent data?
-student t-test
Test for interval three+ groups of independent data?
-ANOVA
+compares against a single dependent variable
-MANOVA
+compares against multiple DVs
What test would you run for interval data with 3+ groups of independent data with confounders?
-ANCOVA
+single DV
-MANCOVA
+multiple DVs
Are student student t-test a ANOVA interchangeable?
yes when comparing two groups
Test for interval 2 groups of paired/related data?
-paired t-test
Test for interval 3+ groups of paired/related data?
- repeated measures ANOVA
- repeated measures MANOVA
Post hoc test for 3+ group comparisons in interval data?
-Student-Newman-Keul test
-Dunnett test
-Dunn test
-Tukey or Scheffe test
+compare all pairwise comparisons possible
+all groups equal in size
What is the Kappa statistic?
-agreement between evaluators (consistency of decisions and determinations)
What are the different interpretations for Kappa?
+1= observers perfectly classify everyone the same way
0= there is no relationship at all between the observers’ classifications, above the agreement that would be expected by chance
-1= the observers classify everyone exactly the opposite of each other
What is the National Clinical Trials number?
-a # assigned by clinicaltrials.gov once research protocol is submitted prior to study initiation
What is the purpose of the NCT#?
-purpose was to reduce publication bias
What do most clinicians want to know how to do with research papers?
- delineate the differences in study designs and to determine which design is most appropriate for a given research question
- evaluate how study design might impact results
- determine strengths and weaknesses for various study designs
- determine what elements can influence study results (biases and confounders)
- determine the most appropriate statistical test for desired comparisons, based on research question and variable measurements
- determine if author’s conclusions are sound and based on actual study results
What is one folly in trying to assess research?
-authors may neglect to provide lucid and complete descriptions of critical, necessary information
What do healthcare professionals use to assess published medical literature now?
CHECKLISTS
Where can you go to get checklists?
-equator network
What checklists are utilized for interventional studies?
-consort (consolidated standards of reporting trials)
+non-inferiority and equivalence trial
+cluster trials
+pragmatic trials
-prisma (systematic reviews of multiple randomized trials)
What checklist is utilized for observational studies?
-strobe (cohort, case-control, cross-sectional)
+strobe-me
+strega
What checklist is used for a non-randomized study?
-trend (reporting evaluations with non-randomized designs of behavioral and public health interventions)
What is the checklist for tumor marker prognostic studies?
-remark (tumor marker prognostic studies)
Checklist for genetic risk prediction studies?
-grips(genetic risk prediction studies)
Checklist for diagnostic studies?
- stard
- quadas-2 (systematic reviews of multiple diagnostic studies)