Final Flashcards
(120 cards)
1
Q
Logic most closely associated with the scientific method
A
Abductive
2
Q
What best defines the scientific method?
A
Proposes useful or predictive claims about the world and tests them by experience
3
Q
Hallmark global research study of the scientific method
A
Interventional study
4
Q
What does random allocation in large research studies minimize?
A
Confounders
5
Q
Least likely to be a part of exploratory data analysis
A
Confidence intervals
6
Q
2 major types of data analysis in quantitative research studies: confirmatory and exploratory. What best characterizes the aim of exploratory data analysis?
A
Present the data in open-minded ways that reveal what the data might mean
7
Q
Probabilistic thinking comes into play in biomedicine most frequently when?
A
Two or more factors have a conditional relationship
8
Q
2 major concepts commonly used in biomedical statistics
A
Central tendency and error analysis
9
Q
Research design is biased under what conditions? (Review)
A
Design systematically favors one or some study outcomes over other outcomes
10
Q
Generally the most informative measure of error in a data set
A
Confidence interval
11
Q
Observational research studies lead to:
A
Correlations
12
Q
What makes interventional research studies so powerful?
A
Order-dependence of intervention and outcome
13
Q
Hypothesis-driven research cannot be?
A
Qualitative
14
Q
Common passageway from qualitative to quantitative studies:
A
Propositions from qualitative study are refined into hypotheses for quantitative study
15
Q
Hypothesis generation is best characterized by
A
Imaginative combining of thoughts to explain a problem
16
Q
What makes a hypothesis potentially innovative rather than simply new?
A
Hypothesis supported by test outcomes has implications far beyond originating field
17
Q
Simply pointing out a high correlation between selected factors are too frequently interpreted to be in a cause and effect relationship. This can be due to our minds being predisposed to?
A
Predict the future given the presence of one of the two factors in question
18
Q
Factors A and B are known to occur independently of each other in each of the 50 US states. You observe factors A and B to be highly correlated in California but independent of each other in Iowa. Which of the following would be the most likely explanation for your observations?
A
A and B in California are each caused by C, but C is absent in Iowa
19
Q
In interventional studies, the intervention is presumed to be…?
A
Cause
20
Q
The best clinical research design is the Randomized Control Trial (RCT). At the heart of RCTs is an intervention and an outcome of interest. What is the logical role of the intervention in RCTs?
A
Presence of the intervention is a predictor of the outcome
21
Q
Baseline rate is determined in what design part of a research study? (Review)
A
Control
22
Q
The gold standard in research design = Randomized Control Trial (RCT). The design features to real with confounding factors that might contribute to the outcome. In an RCT, what gets randomly allocated?
A
Intervention & the control
23
Q
Synonyms for causes and effects in research studies
A
Explanatory & response variables
24
Q
At best, the findings of an observational study are
A
Necessary but not sufficient to establish causal relationships
25
What is a statistical factor altering precision in parameter measures?
Stochastic noise
26
Central tendency = a statistical analysis that determines a single or point value as representative of an entire data distribution. What is a advantage and a disadvantage of the mean versus the median as a measure of central tendency of a data set?
Easily calculated as arithmetic average and more sensitive to extreme values
27
A great way to visualize both the central tendency and spread in a data set?
Five number summary
28
Research studies are complex with lots of design, performance, and outcome parts. In published biomedical research studies, what is ultimately the most important outcome?
Effect size
29
What is a statistic to analyze variation in measures of 2 categorical variables?
Chi-square
30
All research except qualitative studies generates results on a single group or multiple groups of events, objects, or subjects that can be quantified. When analyzing sets of quantitative data, what is usually the most important?
Differences between means
31
Statistical analysis of a categorical study of two groups needs how many cell counts of data?
4
32
Sensitivity and specificity are key concepts in understanding false discovery rates in screening tests. These tests can be in the context of basic science assays or clinical diagnostic tests. Which concept refers to true positives and which concept to true negatives?
Sensitivity refers to the true positives and specificity refers to the true negatives
33
What do a coin toss, dice throw, and dealt cards from a shuffled deck share?
The specific outcome is unknowable
34
Probabilistic thinking is very important to evidence-based medicine on an everyday level. This way of decision making relies on the strength to which two or more features are correlated or independent of one another. If 2 independent events each has a probability of 0.7, what is most likely?
Both events will occur together less than half the time
35
Student’s t-test is a simple statistical test to determine the likelihood that...?
Random variability linked to your measures can account for your research study results
36
Statistical tests are not generally intuitive which leads to a lot of confusion as to what these tests tell us about research measures. The concept and meaning of values of p-values are not well understood generally, including by investigators who use p-values routinely, but must be if we are to decrease the rate at which we publish research studies that turn out to be false. A student’s t-test in a study giving a p-value of 0.001 means...
Difference between means has a 1 in 1000 chance of being random variability
37
What best conveys the magnitude of the outcome of interest important to the research study?
Mean differences or correlations
38
What concept is now used to indicate how likely the mean is what you think it is? (Review)
Confidence interval
39
What combined with the p-value can indicate how likely your hypothesis is false?
Prior probability from other studies
40
What accounts for the assessment that most claimed research findings are false?
Most experiments are under-powered and p-values acceptable for publication too low/high(?)
41
What do interventional studies research?
Cause of an outcome of interest
42
Aim of an intervention research design is consistent with...?
When a factor is removed from the study, the outcome goes away
43
Critical advantage of an RCT study over all other designs is...?
Confounding factors influence on outcome differences across groups is minimized
44
Exclusion and inclusion criteria aim to optimize the possibility that...?
Intervention and control groups begin with the same prognosis
45
It’s often key to design and perform case-control studies so...?
Past conditions of exposure to harm are similar for cases and control subjects
46
A major limitation of case-control studies of decades-long exposure harm
Recall bias with subjects
47
What distinguishes a cohort from a case-control study?
Outcome is absent at the beginning of a cohort study
48
In a cohort study design, what are the minimal cohorts and their outcomes?
Exposed and not exposed at start and disease and not disease at end
49
Which of the following is methodologically similar to basic science hypothesis testing?
Diagnostic testing
50
In diagnostic testing, sensitivity and specificity refer to...?
Fraction of sick people correctly identified as having the condition and the fraction of healthy people correctly identified as not having it, respectively
51
Which medical topic is inappropriate to investigate by randomized controlled studies? (Review)
Harm exposure
52
2 most important endpoint questions for studies about therapy
How large was the treatment effect and how precisely it was measured
53
In prognostic research studies, what issue comes into play for disease whose onset have long latencies?
Bias due to subjects lost to follow up
54
In research studies on prognosis, what is a major goal?
Help clinicians make the right treatment decisions
55
Major difference between a systematic review and a meta-analysis
Data from different studies are statistically aggregated in meta analysis
56
Which of the following biomedical concepts are unique to meta-analyses?
Independent study results are combined, converged, or both to advance findings
57
What part of a study design should include the cause of the biomedical process investigated?
At risk group
58
FINER framework for creating research questions
```
Feasible
Interesting
Novel
Ethical
Relevant
```
59
PICOT framework for creating research questions
```
Patient/problem/population
Intervention
Comparison
Outcomes
Time
```
60
What is excluded from qualitative study designs?
Variables
61
Major factors in study designs providing strong evidence in support of a hypothesis
Temporality and dose response relationships between the candidate cause factor and the candidate outcome measure
62
Student’s t-test is a relatively simple statistical test to determine the likelihood that...
Random variability linked to your measures accounts for your research study results
63
P- values
- Are calculated based on the assumptions that the null is true for the population and that the difference in the sample is caused entirely by random chance => p-values can’t tell you the probability that the null is true or false bc p-value is the probability that the null hypothesis is 100% true
- Even low p-values cannot exclude the null hypothesis in a absolute sense. While a low p-value indicates that your data are unlikely assuming a true null, assuming a true null, it cannot evaluate which of the two competing causes* is more likely:
* the null is true but your sample was unusual
* the null is false
- Medical literature in general is more and more emphasizing confidence intervals and effect sizes over p-values!
64
2x2 table concepts
- Sensitivity: the test’s ability to identify true positives
TPR = TP/P = TP/(TP+FN)
- Specificity: the test’s ability to identify true negatives
TNR = TN/N = TN/(TN+FP)
- Accuracy: combines above as follows
Accuracy = TP+TN/(TP+FN+TN+FP)
- Positive predictive value of test is how likely a positive from this specific test going to actually be a true positive
PPV = TP/(TP+FP)
65
In scientific communications, the rationale of the study is?
Often assumed by authors & not easily grasped by non-experts in the field
66
What best conveys the magnitude of the outcome of interest important to the research study?
Mean differences or correlation coefficient
67
Hypothesis generation classically is best defined by which of the following?
Preparation of precedence, incubation, and illumination
68
Broad goal of an interventional study
Test for cause and effect relationships between measured outcomes and their causes
69
5 numbers of box plot often featured in exploratory data analysis
Minimum, maximum, median, first quartile, and third quartile of a data
70
What part of research study data results in probabilistic thinking being necessary?
The outcome being measured can be conditionally dependent on the independent variable
71
Example of a biased sample in a research study
You set out to find whether all people like you and you report on only the people who do
72
What kind of research is better done using an observational study design?
Research on a topic whose investigation by an interventional study would be unethical
73
What broadly distinguished qualitative, opposes to hypothesis-driven studies?
Discovers, describes, and aims to understand psycho-social dynamics and meanings
74
What makes a question a scientific hypothesis?
Testable by experience in a way that it is possible to be excluded
75
Taking an exploratory data analysis approach to see if there’s any pattern of association between values from 2 parameters from an observational research study, you plot the values on the scatter plot. What are the most important questions to ask when looking at all scatter plots?
Do the data points linearly align? If so, is the line sloping upward or downward?
76
To provide sufficient evidence supporting a cause and effect relationship between two quantitative variables, values of the outcome of interest variable is measured under what conditions of the candidate cause variable?
Comparative conditions of absence, varied amount, concentrations, or frequencies
77
Baseline rates are important to understand whenever doing an interventional study bc the rate might be contributing to the measured effect. When should you control for baseline rate effects by making multiple outcome measures in the absence of the intervention?
In parallel with and before and after the intervention if applied
78
You are reading a study that uses an observational study design to look at how feeding school children big breakfasts lowers obesity. The study looked at households in New York City only surveying the weight of school children and the size in calories of their breakfasts over a five-year period. The study’s chief finding made headlines because it was surprising and seems counter-intuitive. Which of the following about the study is most likely to be true?
Study is an observational study and therefore on its own the findings of cause are flawed
79
You are the analyst collecting data from 100 international hospital sites each studying the exponential decay of tumor mass after applying a promising anti-tumor immunotherapy by plotting the size of the tumor by invasive imaging each week over the course of the six month study. Each site reports weekly to you the mean size of 100 or more of its patients’ tumors, with each patient’s tumor normalized to the initial size immediately before the drug intervention. Each week, the means of the fractional size of the tumors from the 100 hospital sites should show what kind of frequency distribution?
Normal
80
Robustness is an important concept in EDA and descriptive statistics referring to measures of central tendency and spread in a data set or in multiple data sets being compared and contrasted. Which of the following are the most complete and robust measures characterizing data sets?
Median and inter-quartile range
81
A data set of values of a single qualitative variable provides what kind of information?
Frequency of cases in any of two or more categories
82
Two by two frequency tables can be used to compare?
2 categorical variables
83
The fact that most diseases are rare tells you that:
A positive diagnostic test result is likely to be false
84
Student’s t-test is a statistic that is one of the most common procedures in science and medicine. The t-test is the most common statistic of significance used when comparing the mean of two samples of quantitative, normally distributed data sets. The t-test provides statistical analysis of the difference in sample means normalized to the standard error of difference of sample means. The numerator relates to the explanatory variable, whereas the denominator relates to the random noise in the study results. The t-test is special and distinct from other statistics, in particular as opposed to the z-statistic or other statistics, because the t-statistic is valid for small samples. What is the n number of independent samples in a study where the t-test should be used to analyze the probability of the difference between the means of two samples being due to chance alone?
4-30
85
When presenting figures showing analysis of your data, the 95% CI is preferred over the range, SD, and SE assessing your data values because:
CI is an easily understood indicator of how precise the mean of the values was determined
86
What does the ‘p’ in p-value stand for and the numeric value indicate?What does the ‘p’ in p-value stand for and the numeric value indicate?
‘Probability’ of random variation accounting for the difference observed in test versus control cases
87
What are likely the best sources for evidence supporting best clinical practice and what makes them so?
Biomedical journals reviewed by fellow clinicians and scientists published based on strength of evidence and reasoning
88
Without a well-formulated question, we are likely to search data bases for:
- Hours or even days
- The too good to be true
- Irrelevant facts and figures
- Wherever it may lead, and more
89
The time and effort spent searching data bases for biomedical articles can be cut ten-fold by first choosing to search based on:
Specific kind of clinical article specified by descriptors of the relevant type of clinical question
90
At the stage of analyzing and evaluating the one or the few articles that survived your triage in Critical Appraisal, answering facets of your major formulated question and associated questions can be easier, faster, and more effective, taking advantage of the fact that:
Answers to the types of questions you will need to answer are located in typical sections of the article
91
What would have been the Odds Ratio if the women in Europe aged 50-80 years on the combination drug therapy with hip fractures were 100 and without hip fractures were 1500, and what does that tell you about the influence of potential inaccuracy in the recording or reporting of the presence or absence of hip fractures due to recall bias or confounders?
Use of the combination drug therapy reduced the likelihood of hip fracture by 33% and the likelihood is very sensitive to misclassification of rather few subjects compared to the total subjects in the study
92
What is the major distinction making Cohort studies at least by design better than Case-Control studies?
Cohort studies are prospective with participants not having the disease or medical condition at the onset of the study
93
The Scientific Method is a strategy for generating new knowledge about the universe and features as part of its process tests by experience. You have a claim about say how the body works and you test that claim in the real world. If the results of the test exclude the claim, you discard it and find another. If the results of the test support the claim, you continue to use it and test it. In this process, what are the major cognitive steps?
Imaginative hypothesis generation and critical reasoning
94
Medical outcomes are generally far more complex than those from basic science experiments. Large drug trials, for example, have a number of variables involving everything from the design of the trial to what often can be a wide range of differences in the patients enrolled. When comparing the outcome of a drug trial on patient groups, what can be difficult or impossible to know yet alters outcome measures?
Confounding factors
95
Unlike Confirmatory Data Analysis (Statistical Hypothesis Testing), Exploratory Data Analysis embraces the 3 principles of critical reasoning. What principles are embraced by the Exploratory Data Analysis?
Analyzing the distribution of the data while suspending judgment as to the conclusion
96
When one thinks of statistics in research studies, one often thinks of p-values. After all, they are nearly ubiquitous in published studies. The p-value quantifies the probability that the data in question can be the result of random variability alone. It doesn’t at all address the formulated research question or hypothesis being investigated. What could be combined with the p-value and would assess the likelihood of the hypothesis being tested in one’s research study?
Prior probability calculated from previous experiments or observations related to study
97
When teachers and students are asked what is critical thinking, most typically, they go on for many minutes describing it. In a few words, how would you define critical thinking or critical reasoning?
Analyzing, evaluating, & integrating
98
A common source of problems at the level of design originates from investigators using a “convenience” sample. This is a strategy to generate so that the data are robust. However, typically, these studies take the form of first presenting an issue of broad appeal then giving people in a high traffic area (popular internet site or place on campus) the option to take the survey. What is the likely source of bias in this approach?
People who are highly opinionated will be over-represented in the survey results
99
In Study I, subjects were exposed to a naturally occurring environmental factor that is thought to block development of disease X in region I. The factor is absent from region II where disease X has its usual high prevalence. Study II was in region II where the isolated factor was given or not as interventions and disease X monitored. Given limited resources, what study design feature would cause you to choose Study I or Study II to determine whether the environmental factor prevents disease X?
Exposure of factor to a given subject is known and potential effect measured shortly after
100
Francis Bacon, often referred to as a pioneering philosopher of the scientific method, did much to define its underlying logic. However, which of the following was the one most critical component that Bacon said little if anything about?
Interleaving of free imagination with coherent reasoning on the object of investigation
101
What features are the most important advantage of interventional designs over observational design in research studies?
Multiple comparator groups to determine primary intervention effectiveness (?)
102
What are major components to a logical argument that results from applying the scientific method?
Evidence from testing the hypothesis in experience
103
Center stage in quantitative research are measures of how dispersed values in a data set are from its average value. Dispersion from the average value tells you about:
Confidence in its precision and likely distinctness from the average value of other data sets
104
What are major advantages of using graphs that display box plots of a quantitative outcome for each of multiple qualitative explanatory variables in a quantitative research study?
Central tendency in the median and spread in interquartile range for each data set and visual comparisons and contrasts of medians for effect sizes and interquartile range for preciseness of measures across all explanatory factor variables
105
What distinguishes qualitative research and qualitative variables?
Qualitative research has no numbers that are mathematically meaningful. Qualitative variable data has the number of cases, which are mathematically meaningful, which are scored for each category of the qualitative variable
106
The first and most important step in scientific thinking, in contrast to pedestrian thinking, is to test whether your explanatory factor variable and your outcome variable are coincidental. Think of it as testing for the null hypothesis for causality in quantitative research. What do you need to test for independence in your data?
All cases when both, the one or the other, or none of your variables were present
107
What are the reasons why a patient who tests for positive for a disease does not necessarily have a 100% probability of actually having the disease?
Diagnostic tests have a significant rate of false positives
108
The student’s t-test boils down to a ratio of what statistical variables over what other statistical variables?
Effect size normalized by a weighted average of experimental error
109
What is a strength and a weakness inherent with the one-way ANOVA statistic?
Allows statistical comparisons of two or more data sets but it does not itself indicate which pair(s) of data sets are likely to have real differences in means
110
What are the advantages of presenting estimates of variability of a research study by 95% confidence intervals rather than p-values?
CI’s are straightforward visual and conceptual relationship to the point estimate of the central tendency of the data set(s)
111
Why do physicians make sure that they know the epidemiology of the diseases in his/her area of practice?
A centerpiece of epidemiology is the prevalence, which is important in properly interpreting diagnostic test results
112
When reading the primary biomedical literature, physicians & scientists should be?
Skeptical
113
As a future clinician, scientist, or clinician-scientist, how many articles do you think will be published in your area of specialization per week?
100 or 1000(?)
114
Given the deluge of daly publications in biomedicine, what will be your primary strategies to ensure that you learn important advances in your field for your patients and research?
Structured approach based on a structured question and study designs and types appropriate to that question
115
What are major divide and conquer strategies associated with critical appraisal of the biomedical literature?
Investigate the type of research designs & study designs that are most relevant to your structured question
116
What are over-arching topics to analyze and evaluate in any grant, publication or scientific dialogue?
Claim, Rationale, Evidence, Alternatives, Meanings
117
After triaging the medical literature based on your PICOT question, while investigating the full article, what are you focusing on for evaluating the strength of the findings?
Effect sizes and associated confidence intervals
118
When surrogate endpoints are used, which of the following properties might they have in an RCT?
- Cost far less to measure compared to measuring the actual clinical outcome in the patient
- Be an early change to allow for shorter trial times than the actual patient outcome measure
- Exhibit dose-response relationship between the intervention and the surrogate outcome measure
- Be more amenable to blinding compared to measuring the actual clinical outcome in the patient
119
What feature distinguishes a cohort from a case-control study in such a way that makes the strength of evidence to be stronger by design in the former, compared to the latter study type?
Cohort studies are prospective, whereas case-control studies are retrospective
120
What concepts are important in evaluating evidence for cause and association, over coincidence?
- Strength of evidence, enormous increase in outcome in presence but not is absence of candidate cause
- Reproducibility, under widely differing circumstances, observed repeatedly, k independently by different investigators, in different places, circumstances, and times
- Specificity, of the association, limited to certain subjects or patients, to particular sizes, to certain types of disease, with no association between these and other modes of dying
- Temporality, key, the candidate cause evident must occur before the outcome of interest, particularly important concept for diseases and conditions with long, years or decades latency
- Biological gradient, or dose-dependent relationship between can dilate cause and outcome measure of interest
- Biomedical plausibility and coherence with the framework of principles and facts known in the specific discipline or field
- Intervention, when possible to directly test the candidate cause and measure the outcome of interest by studies designed to test for the concepts of association and causality above
