Module 2: Study Designs and Sampling Flashcards
All sampling units are selectable
-every sampling unit in the statistical population must have some non-zero probability of being included in your sample (if there is both brown and black penguins and you state that the statistical population is all the penguins in the colony, then every penguin in the colony must have a chance of being included)
Selection is unbiased
-avoid bias so that sample is a representation of larger statistical population, avoid using senses to select sampling units (don’t select only baby penguins because they are easier to catch)
Selection is independent
-selection of one sampling unit cannot influence the probability that any other particular sampling unit is selected (can not interview anyone in same group)
All samples are possible
-all samples that could be created from the statistical population are possible (if u are surveying both east and west sides of the city, composition must include sampling units from both the east and west sides)
Observational studies
-based on observations from a statistical population, different from an experimental study because researcher has no control over the variables
Goals and limitations of observational studies
-goals: characterize something about an existing population
-drawback: can provide associations but cannot provide causation conclusions
Confounding variables
-uses explanatory variables (smoking) and response variable (lung cancer), but also includes confounding variables (age etc)
Simple random survey (observational study design)
-identifies every sampling unit in the statistical population and then selecting a random subset of those to be in the sample (create list and randomly draw from it)
Stratified survey (observational study design)
-when there are subgroups in statistical population that can influence the study results (if we wanted to estimate wages of people across canada, the data would be dominated by people in ontario, so we break population into strata with different weighting)
Cluster survery (observational study design)
-create groups where the non-relevent heterogenity is contained within each group (children earn less than adults so if you are measuring household earnings you should sample each house, not each person)
Case-control survey (observational study design)
-used to compare data between 2 groups, there is both a case and control group and the control group contains sampling units without the response variable of the case group
Cohort survey (observational study design)
-follow sampling units over time, looking for the development of a particular response variable (outcome is unknown when the sampling units are selected)
Retrospective vs prospective
-retrospective: studies where the outcome is already known
-prospective: studies where the outcome is not yet known
Cross-sectional vs longitudinal
-cross-sectional: studies that the response variable are only taken at a single snapshot in time
-longitudinal: studies that the response variable is recorded at multiple points in time
Experimental study
-goal is to study the effect of one or more manipulated variables on one or more response variables, by manipulating variables (are used to establish cause and effect among variables)
Replication
-idea that a treatment will be repeated a number of times to see how repeatable a measured outcome is
Pseudoreplication
-error in the design of an experimental study where the observation units are analysed rather than the sampling units
Control treatment (experimental study)
-reference treatment to compare against the treatment levels that alter the explanatory variable
Blocking (experimental study)
-analogous to stratified sampling but for experimental studies, it is used to control for variation among sampling units
Blinded (experimental study)
-sampling unit alone (single blind), and sometimes also the researcher (double blind) does not know what treatment they are assigned to
Placebo (experimental study)
-used for control treatment to help accomplish a blinded design, is a substance or treatment that has no affect on the response variable
Sham treatment (experimental study)
-like gold standard drug/treatment