Observational Study
Based on data in which no manipulation of factors has been employed
The researchers merely observe what is happening or what happened in the past
Two types: Retrospective and Prospective
Not possible for either type to demonstrate a casual relationship
Retrospective Study
An observational study
Subjects are selected and then their previous conditions/behaviors are determined
ex: first identified subjects who studied music and then collected data on their past grades
Does not usually need to be based on random sampling
Usually focuses on estimating differences between groups or associations between variables
Can have errors because these are based on historical data
someone could not fully remember etc.
woah that so retro and old
Prospective Study
An observational study
Subjects are followed to observe future outcomes
Identifying subjects in advance and collecting data as events unfold
Again, since this is a type of observational study, no treatments are deliberately applied
Usually focuses on estimating differences among groups that might appear as the groups are followed during the course of the study
Experiment
Manipulated factor levels to create treatments
Randomly assigns subjects to these treatment levels, then compares the responses of the subject groups across treatment levels
Random Assignment
To be valid, an experiment must assign experimental units to treatment groups at random
Factor
An explanatory variable whose levels are manipul;aged by the experimenter
Experiments attempt to discover the effects that differences in factor levels may have on the responses of the experimental units, randomly assigned to subjects
factor = explanatory variable
factory ends in a y and what comes before y? x-explanatory
Response Variable
Values are compared across different treatments
In a randomized experiment, large response differences can be attributed to the effect of differences in treatment level
Experimental Units
Individuals on whom an experiment is performed
ex: rats, days, petri dishes of bacteria
Subjects/Participants
When the individuals on whom an experiment is performed are human they are called…
Level
The specific values that the experimenter chooses for a factor
Called the “levels of the factor”
Treatment
The combination of specific levels from all of the factors r that an experimental unit receives
different levels of a single factor
OR
combinations of levels of two or more factors
Principles of Experimental Design
Control
Randomize
Replicate
Block
Control
It is imperative that an experiment…
controls sources of variation (other than the factors we are testing) by making conditions as similar as possible for all of the treatment groups
o get and find randomization
Randomize
It is imperative that an experiment…
randomizes/equalizes the effects of unknown or uncontrollable sources of variation
How you will randomize it, be specific
ex: randomly assign dog #’s 1-14 etc.
OR
list of dogs and randomly assign groups 1-4, stop after 15 groups, 1’s 2’s 3’s 4’s…
Replicate
It is imperative that an experiment…
replicates, because it often happens that subjects of the experiment are not a representative sample from the population of interest aka drawing conclusions about the world is impossible unless we replicate/repeat our results
Ways to implement replication in an experiment:
Apply each treatment to a large number of subjects/objects within one experiment
because the outcome of an experiment on a single subject is an anecdote, not data.
ex: 90 people within the study
OR
The entire experiment is repeated on a different population of experimental units
Block/Blocking
To reduce the effects of identifiable attributes of the subjects (that may affect their responses but can’t be controlled)
If we group similar individuals together
and then randomize treatment assignments within each of these blocks
we can account for much of the variability due to the differences among the blocks
so we can see the differences caused by the treatments more clearly
ex: you could block by dog breed, by current weight of the dog, or age of the dog (harder to lose weight as they get older)
Identify what characteristic will have the biggest impact and then decide to block
What is an important compromise between randomization and control
Blocking!
Unlike the other 3 principles of an experiment, which fourth principle is not required in an experiment
Blocking
Randomized Block Design
an experimental design where subjects or experimental units are grouped into blocks, with the different treatments to be tested randomly assigned to the units in each block
The randomization of an experiment occurs only within the blocks
ex: plants from one store can’t be assigned to treatment groups for the other store plant group
Completely Randomized Design
All experimental units have an equal chance of receiving any treatment
Statistically Significant
When an observed difference is too large for us to believe that it is likely to have occurred by chance (this difference was not due to chance)
Therefore the difference must be due to the treatments and is statistically significant!
Bigger picture of types of data collection
Sample surveys try to estimate population parameters
whereas
experiments try to assess the effects of treatments
Control Group
The experimental units assigned to the control treatment (a baseline, default, null, placebo treatment)
The control group’s responses provide a basis for comparison
Blinding
Any individual associated with an experiment who is not aware of how subjects have been allocated to treatment groups is said to be blinded
Used to combat a possible source of confounding
Two main classes of individuals who can affect the outcome of an experiment:
Those who can influence the results
subjects, treatment administrators, etc.
Those who evaluate the results
judges, treating physicians, etc.
Single-Blind
When every individual in either of these classes (those who can influence the results and those who can evaluate the results) is blinded
Double-Blind
When everyone in both classes(those who can influence the results and those who can evaluate the results) is blinded
Placebo
A treatment known to have no affect
Administered to one group so that all groups experience the same conditions
Placebo Effect
The tendency of many human subjects (20% or more) to show a response even when administered a placebo
Only by comparing with a placebo can we be sure that the observed effect of a treatment is not simply due to the placebo effect
ex: a person struggling with insomnia takes a placebo (sugar pill) but instantly falls asleep because they believe they are receiving a sleep aid like Ambien or Lunesta.
Matching
In an observational study
Subjects who are similar in ways not under study may be matched and then compared with each other on the variables of interest
Like blocking, matching can be used to reduce unwanted variation/variability
Confounding
BAD!!
When the levels of one factor are associated with he levels of another factor in such a way that their effect can’t be separated
Think: like pouring two miscible liquids together, they cant be secreted from one another
This is just like how the effects of two confounded explanatory variables can’t be separated either
Can occur from a two-factor combination as well
Lurking variable v. Confounding Variable
Lurking variable: associated with both x and y, make it appear as though x is causing y, causes the other two variables to have a relationship
Confounding variable: We can’t tell whether any effects as caused by out factor, by the confounding variable, or by both working together, affects the response variable
In summary, the best experiments are usually
randomized
comparator el
double-blind
placebo-controlled
Matched Pairs Experiment
A special case of randomized block design, where an experiment only has two treatment conditions. The participants are grouped together into pairs based on an equivalent variable, such as age or gender. Within each pair, subjects are randomly assigned to one of two treatments.