Exam 3 Flashcards
Data will be collected on desired ________, both dependent (outcome) and independent.
Variables
The 3 primary levels (groupings) for variables (data) are…
1) Nominal
2) Ordinal
3) Interval or Ratio
The 3 primary levels are based on 3 key attributes, which are…
1) Order/Magnitude
2) Consistency of scale/ Equal distances
3) Rational absolute zero
This primary level has NO order/magnitude, NO consistency of scale/equal distances, and is based on named categories. Dichotomous/binary; Non-ranked.
Nominal
If something has order/magnitude, but only has 2 categories, then it is ALWAYS (NOMINAL/ORDINAL).
Nominal
This primary level has order/magnitude, but NO consistency of scale/equal distances. It is ordered, rank-able categories.
Ordinal
This primary level has order/magnitude and equal distances (units).
Interval/Ratio
(INTERVAL/RATIO) has an arbitrary zero value (0 doesn’t mean absence, it can go negative); (INTERVAL/RATIO) has an absolute zero value (0 means absence of measurement value, it can NOT go negative).
Interval; Ratio
Which primary levels are considered ‘discrete’ or ‘continuous’?
Nominal and Ordinal = Discrete
Interval/Ratio = Continuous
ALL statistical _____ are selected based on the level of data being compared.
Tests
After data is collected, we can appropriately go (UP/DOWN) in specificity/detail of data measurement levels, but we can never go (UP/DOWN).
Down; Up
________ statistics are non-comparative, simple description of various elements of the study’s data.
Descriptive
Examples of descriptive statistics, measures of central tendency (dispersion or spread) are _____/_____/_____; ______/______/______; and ________ ________.
Mode/Median/Mean
Minimum/Maximum/Range
Interquartile Range (IQR)
Interquartile range is the middle ___% of the data.
50%
The average of the squared-differences in each individual measurement value (x) and the groups’ mean is called…
Variance
The square root of the variance value (restores units of mean) is called…
Standard Deviation (SD)
Stats tests useful for normally-distributed data are called _________ tests.
Parametric
_______ representations shows the shape of data (i.e., bell curve).
Graphical
Normally distributed data is considered (SYMMETRICAL/ASYMMETRICAL).
Symmetrical
What a dataset is normally-distributed the following values (parameters) are equal/near equal…
Mean and Median
A standard deviation of +/- 1 from the mean is ____%.
68%
A standard deviation of +/- 2 from the mean is ____%.
95%
A standard deviation of +/- 3 from the mean is ____%.
99.7%
A (POSITVELY/NEGATVELY) skewed graph is when the mean is higher than the median, and the tail (on graph) points to the right.
Positively
Skewed data distribution is considered (SYMMETRICAL/ASYMMETRICAL), with one ‘tail’ longer than another on the graph. A distribution is skewed anytime the median differs from the mean.
Asymmetrical
A (POSITIVELY/NEGATIVELY) skewed graph is the when the mean is lower than the median, and the tail (on graph) points to the left.
Negatively
This is a measure of the asymmetry of a distribution. The perfectly-normal distribution is symmetric and would have a value of 0 with this measurement.
Skewness
T/F. A non-zero skewness value DOES NOT mean it’s skewed.
True
This is a measure of the extent to which observations cluster around the mean. For a normal distribution, the value is 0.
Kurtosis
A positive kurtosis means there is (MORE/LESS) cluster, and a negative kurtosis means there is (MORE/LESS) cluster.
More; Less
What are the 3 required assumptions of interval/ratio data for proper selection of a parametric test?
1) Normally-distributed
2) Equal variances
3) Randomly-derived and Independent
What test is used to assess for equal variances between groups?
Levene’s Test
How is interval data handled that is NOT normally-distributed?
1) Use statistical test that does not require it to be normally-distributed
(OR)
2) Transform data to a standardized value (hoping this makes data normally-distributed)
This is a research perspective that states there will be NO true difference between the groups being compared. It is the most conservative and commonly utilized.
Null Hypothesis (H0)
This type of error is NOT accepting the null hypothesis when it is actually true, and you should have accepted it. There really was no true difference between groups but you believed there was a difference.
Type 1 Error (Alpha)
This type of error is accepting the null hypothesis when it is actually false, and you should NOT have accepted it. There really is a true difference between groups but you believed there was no difference.
Type 2 Error (Beta)
This is the statistical ability of a study to detect the true difference, IF one truly exists between group-comparisons, and therefor the level of accuracy in correctly accepting/not accepting the null hypothesis.
Power (1-Beta)
T/F. The larger the sample size, the greater the likelihood (ability) of detecting a difference if one truly exists.
True
The larger the sample size, then there is a(n) (INCREASE/DECREASE) in power.
Increase
We accept a ____% risk of making Type 2 errors (___% power).
20%; 80%
We accept a ____% risk of making Type 1 errors (____% power).
5%; 95%
The 3 things that determine sample size are –
1) (MINIMUM/MAXIMUM) difference between groups deemed significant
2) Expected variation of measurement
3) Type 1 (alpha) and Type 2 (beta) error rates and confidence interval (ranges from 90%-99%
Minimum
The (SMALLER/LARGER) the difference between groups necessary to considered “significant”, the greater the sample size is needed.
Smaller
If the p-value is lower than the pre-selected alpha (Type 1) value (customarily 5% (0.05)) then we say it (IS/IS NOT) statistically significant.
Is
If the p-value is less than the alpha percentage-risk error (5%), then we (REJECT/ACCEPT) the null hypothesis.
Reject
Interpretation of a pre-set (a priori) p-value —- The probability of making a Type 1 error if the null hypothesis is (ACCEPTED/REJECTED).
Rejected
If we are interpreting the p-value for more than 2 categories, then we focus on the ______ and ______ values to prove a difference.
Lowest; Highest
The 2 times we want there to be NO statistical difference (accept null hypothesis) is when there is —
1) ________ data - starting out
2) In a _______ test
Baseline; Levene’s
If the p-value is 0.91 and you say there IS a significant difference, then you have a ____% chance of committing a Type 1 error.
91%
When groups are equal, the p-value is ____.
1
In interval data, we use the Levene’s test to assess for equal variances between groups. If there IS a difference (not equal variances) then we do not use the next test, called the _______.
T-Test
If you don’t use the same ________ words for confidence interval (CI), then it is not statistically significant.
Directional
Journals are moving away from solely reporting ______, or even showing them at all.
P-values
If a confidence interval crosses ____ for ratios (RR/OR/HR) or _____ for absolute differences, then they are NOT significant.
1.0; 0.0
P-values can only tell us if data is statistically significant, but _______ _______ can tell us if data is statistically significant and the range group where the difference might lie.
Confidence Interval (CI)
When reviewing the findings of a study, what question should ALWAYS be asked?
Does “statistical” significance actually confer meaningful, “clinical” significance?
What is the 1st question that must be asked when selecting the correct statistical test?
What data level is being recorded? (nominal, ordinal or interval)
What is the 2nd question that must be asked when selecting the correct statistical test?
What type of comparison/assessment is desired?
This type of statistical test provides a quantitative measure of the strength and direction of a relationship between variables. Values range from -1.0 to +1.0
Correlation test
A ______ correlation is a correlation that controls for confounding variables.
Partial
When there is a positive correlation of exactly +1.0, then the graph will display a ____ degree line in the positive direction.
45
When there is a negative correlation of exactly -1.0, then the graph will display a ____ degree line in the negative direction.
45
What is the name for a nominal correlation test?
Contingency coefficient
What is the name for an ordinal correlation test?
Spearman correlation
What is the name for an interval correlation test?
Pearson correlation
In the interval correlation test (Pearson correlation) a p value > 0.05 means there is no ________ correlation. There may still be ________ correlations present.
Linear; Non-linear
This type of statistical test compares the proportion of events over time, or time-to events, between groups.
Survival test
What can all survival tests be represented by graphically?
Kaplan-Meier curve
What is the name of the nominal survival test?
Log-Rank test
What is the name of the ordinal survival test?
Cox-Proportional Hazards test
What is the name of the interval survival test?
Kaplan-Meier test
This type of statistical test provides a measure of the relationship between variables by allowing the prediction about the dependent, or outcome, variable knowing the value/category of independent variables. Also able to calculate OR for a measure of association.
Regression test
What is the name of a nominal regression test?
Logistic regression
What is the name of an ordinal regression test?
Multinomial Logistic regression
What is the name of an interval regression test?
Linear regression
When the answer to the 2nd question of selecting the correct statistical test is “frequencies/counts/proportions” then there must be 2 follow up questions after (3rd and 4th question), which are…
3) How many groups are being compared?
4) Is the data independent or related (paired)?
If you have 2 groups of independent nominal data, then you use which test?
Pearson’s Chi-square test
If you have 3 or more groups of independent nominal data, then you use which test?
Chi-square test of independence
What type of test has a skewed curve and compares “should have expected” from the real value?
Chi-square test
If you have independent nominal data with 2 or more groups with an expected cell count of less than 5, which test should you use?
Fisher’s Exact test
For statistically significant data (in all data types) of 3 or more comparisons, you have to perform post-hoc testing to determine which tests are different. With nominal data, you use which post-hoc test?
Bonferroni test of inequality (Bonferroni correction)
If you have paired/related nominal data with 2 groups, which test should you use?
McNemar test
If you have paired/related nominal data with 3 or more groups, which test should you use?
Cochran test
If you have 2 groups of independent ordinal data, which test do you use?
Mann-Whitney test
If you have 3 or more groups of independent ordinal data, which test do you use?
Kruskal-Wallis test
If you have 3 or more groups of ordinal data (independent or paired/related) and it is statistically significant, you must do a post-hoc test. Ordinal has 3 possible post-hoc tests, which are…
Student-Newman-Keul test
Dunned test
Dunn test
This ordinal/interval post-hoc test is used to compare all pairwise comparisons possible. All groups must be equal in size.
Student-Newman-Keul test
This ordinal/interval post-hoc test is used to compare all pairwise comparisons against a single control (i.e., center of pinwheel). All groups must be equal in size.
Dunnett test
This ordinal/interval post-hoc test is used to compare all pairwise comparisons possible. Useful when all groups are NOT of equal size. (Good for lost-to-follow-ups)
Dunn test
Ordinal data tests compare the (MEAN/MEDIAN) between groups, while the interval data tests compare the (MEAN/MEDIAN) between groups.
Median; Mean
If you have 2 groups of paired/related ordinal data, then which test is used?
Wilcoxon Signed Rank test
If you have 3 groups of paired/related ordinal data, then which test is used?
Friedman test
If you have 2 groups of independent interval data, which test is used?
Student t-test
If you have 3 or more groups of independent interval data, which test is used?
Analysis of Variance (ANOVA)
If you have 3 or more groups of independent interval data with confounders, which test is used?
Analysis of Co-Variance (ANCOVA)
If you have 2 groups of paired/related interval data, which test is used?
Paired t-test
If you have 3 or more groups of paired/related interval data, which test is used?
Repeated Measures ANOVA (1 DV)
If you have 3 or more groups of paired/related interval data with confounders, which test is used?
Repeated Measures ANCOVA
If you have 3 or more groups of interval data (independent or paired/related) and it is statistically significant, you must do a post-hoc test. Interval has 5 possible post-hoc tests, which are…
Student-Newman-Keul test Dunnett test Dunn test Tukey (or) Scheffe tests Bonferroni correction
This interval post-hoc test compares all pairwise comparisons possible. All groups must be equal in size. One is slightly more conservative than Stu-N-K and the other is less affected by violations in normality and homogeneity of variances, which makes it the most conservative.
Tukey test Scheffe test (most conservative)
This interval post-hoc test adjusts the p-value for the number of comparisons being made. It is very conservative.
Bonferroni correction
This is a type of correlation test showing relationship or agreement between evaluators (consistency of “decisions” or “determinations”).
Kappa statistic
In a kappa interpretation, this means the observers perfectly “classify” everyone exactly the same way.
+1
In a kappa interpretation, this means there is no relationship at all between the observers’ “classifications”, above the agreement would be expected by chance.
0
In a kappa interpretation, this means the observers “classify” everyone exactly the opposite of each other.
-1
Kappa (K) value can be a + or - . + = (GOOD/POOR) agreement, and - = (GOOD/POOR) agreement.
Good; Poor
This is a unique identifier number assigned by clinical trials.gov once research protocol is submitted prior to study initiation. Its purpose is to reduce publication bias.
National Clinical Trials (NCT) number
This gives us a digital number to go find where an article is found on the web. It will take you directly to the article, especially from references.
DOI
This checklist is used for randomized interventional trials/studies.
CONSORT (Consolidated Standards of Reporting Trials)
This checklist is used for systematic reviews of multiple randomized trials (interventional).
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses)
This checklist is used for observational studies (cohort, case-control, cross-sectional).
STROBE (Strengthening the Reporting of Observational Studies in Epidemiology)
This checklist is for molecular epidemiology observational studies.
STROBE-ME
This checklist is for genetic association observational studies.
STREGA (Strengthening the Reporting of Genetic Association studies)
This checklist is for reporting evaluations with non-randomized designs of behavioral and public health interventions.
TREND (Transparent Reporting of Evaluations with Non-randomized Designs)
This checklist is for tumor marker prognostic studies.
REMARK (Reporting Recommendations for Tumor Marker Prognostic studies)
This checklist is for genetic risk prediction studies.
GRIPS (Genetic Risk Prediction Studies)
This checklist is for diagnostic studies.
STARD (Standards for the Reporting of Diagnostic Accuracy studies)
This checklist is for the systematic reviews of multiple diagnostic studies.
QUADAS-2 (Quality Assessment of Studies of Diagnostic Accuracy in Systematic Reviews)
