Applied Statistics Questions Exams 2023, Exams of Statistics

Download Applied Statistics Questions Exams 2023 and more Exams Statistics in PDF only on Docsity! Applied Statistics Questions Exams 2023  counterfactual - what would happen to the same people exposed to a causal factor if they simultaneously were not exposed to the causal factor  can never be realized but is a good model to keep in mind when designing a study to address cause- probing questions  effect - represents the the difference between what actually did happen with the exposure and what would have happened without it.  Criteria for Causality - 1. Temporal- A cause must precede in effect in time  2. Relationship- there must be an empirical relationship between the presumed cause and the presumed effect; must show an association between the 2  3. No confounders- The relationship cannot be explained as being caused by a third variable  4. Coherence involves having similar evidence from multiple sources  5. Consistency involves having similar levels of statistical relationship in several studies  6. Biological plausibility is evidence from laboratory or basic physiologic studies that a causal pathway is credible  EXPERIMENTAL DESIGN -  experiment - aka: randomized control trials (RCTs)  researchers are active agents, not passive observers  isolates phenomenon and controls conditions under which they occur  RCTs are the best possible design for illuminating causal relationships but is not always possible to use for various ethical or practical reasons.  Characterized by: manipulation, control, randomization  Manipulation - the researcher does something to at least some of the participants  an intervention  the control group eventually receives the full experimental intervention after all the research outcomes have been assessed  attention control group - Sometimes researchers include an attention control group when they want to rule out the possibility that an intervention effects are caused by the special attention given to those receiving the intervention rather than by the actual treatment content  Randomization - aka: random assignment or random allocation  involves assigning participants to treatment conditions at random  participants have equal chance of being assigned to either group  no systematic bias in the groups with respect to pre intervention attributes  preferred method of equalizing groups but there is no guarantee that groups will be equal  the risk of unequal groups is high when the sample size is small  Matching - conscious attempt to control the characteristics that are likely to affect the outcome  to match effectively, the researcher must know the characteristics that are likely to affect the outcome  Complete randomization - the most straightforward randomization procedure  to simply allocate each person as they enroll in the study on a random basis  each person has a 50-50 chance of being assigned to the intervention group  problem: large imbalances in group size can occur  Simple randomization - started with a known sample size and then pre specifying the proportion of subjects who will be randomly allocated to different treatment conditions  Table of random numbers - used prior to widespread use of computers  the larger the sample the stronger the likelihood that the groups will be balanced on factors that could affect the outcomes  Random assignment vs Random sampling - Randomization (Random assignment) is a signature of an experimental design.  If there is no random allocation of participants to conditions, than design is not a true experiment.  Random sampling is a method of selecting people for a study.  Random sampling is NOT a signature of experimental design. In fact, most RCTs do not involve random sampling  Allocation concealment - prevents those who enroll from knowing upcoming assignments  intended to prevent biases that could stem from knowledge of allocations before assignments actually occur  Sequentially numbered, opaque sealed envelopes (SNOSE) - contain assignment information  each new participant receives the next envelope in the sequence  can be subject to tapering  Baseline data - pre intervention data  should be collected before randomization  data on the dependant variable are collected only once- after randomization and completion of the intervention  Pretest-posttest design - aka: before-after design  a type of basic experimental design  involves collection of baseline data and post intervention data collections  allows researchers to analyze change  Repeated measures designs - data collection on multiple post intervention points  allows analysis of differences between groups and changes within groups over time  Parallel groups - involve manipulating only one independent variable and randomizing participants to different treatment groups  Factorial design - researchers manipulated 2 or more variable simultaneously  allows researchers to evaluate not only main effects (effects from experimentally manipulated variables) but also interaction effects (effects from combining treatments)  Cells - treatments conditions  -----------------  factors- independent variables in a factorial design  levels- the dimensions of the design  A crossover design - involves exposing the same people to more than one condition  When there are three or more conditions to which participants will be exposed, the procedure of counterbalancing can be used to rule out ordering effect -  Carryover effects - when people are exposed to two different treatments or conditions, they may be influenced in the second intervention by their experience in the first  Washout period - a period of no treatment exposure in between interventions in attempt to prevent carryover effects  Hawthorne effect - a placebo type effect caused by people's expectations  QUASI-EXPERIMENTS -  Quasi-experiments - controlled trials without randomization  involve an intervention but lack randomization  some also lack a control group  The nonequivalent control group pretest-posttest design - involves two groups of participants, for whom outcomes are measured before and after the intervention  this design is weaker because it cannot be assumed that the experimental and comparison groups are initially equivalent  Comparison group - In quasi-experiments, the term comparison group is often used in lieu of control group to refer to the group against which treatment group outcomes are evaluated  Propensity matching - a more sophisticated form of matching  involves the creation of a single propensity score that captures the conditional probability of exposure to a treatment given various pre-intervention characteristics, experimental and comparison groups can then be matched using this score  a correlation is a relationship or association between 2 variables, that is, a tendency for a variation in one variable to be related to variation in another  correlation does not prove causation  retrospective design - Studies with a retrospective design are ones in which a phenomenon existing in the present is linked to phenomena in the past  the signature of retrospective study is that the researcher begins with the dependant variable and then examines whether it is correlated with one or more previously occurring independent variables  Case-control design - a type of retrospective design that looks at groups of people with a certain phenomenon (cases) and compared them to groups of people that do not experience the phenomena (controls)  problem: the 2 groups are almost never totally comparable with respect to all potential factors influencing the outcome  prospective design - In correlational studies with a prospective design, aka: cohort design, researchers start with a presumed cause and then go forward in time to the presumed effect  the strongest design for Prognosis questions and for Etiology questions when randomization is impossible  Natural experiment - one in which a group exposed to a phenomenon with potential health consequences is compared with a non exposed group  considered non experimental because the researcher does not intervene  Path analysis - used to test theories of causation  researchers test a hypothesized causal chain among a set of independent variables, mediating variables and dependant variables  allow researchers to confirm whether nonexperimental data conform sufficiently to the underlying model to justify causal inferences  Descriptive research - their purpose is to observe, describe, and document aspects of a situation as it naturally occurs and sometimes to serve as a starting point for hypothesis generation or theory development  descriptive correlational research - The aim of descriptive correlational research is to describe relationships among variables rather than to support inferences of causality  Univariate studies - not necessarily focused on a single variable; aims to describe the frequency of occurrence of a behavior or condition; involves multiple variables but the primary purpose is to describe the status of each and not to relate them to one another  Prevalence studies - estimate the prevalence rate of some condition at a particular point in time  Incidence studies - estimate the frequency of developing new cases  Relative risk - an estimated risk of "caseness" in one group compared with another  The primary weakness of correlational studies for cause-probing questions is that they can harbor biases, such as self-selection (aka: selection bias) into groups being compared -  Prospective vs. Retrospective - Prospective looks forwards in time  Retrospective looks backward in time