exploratory data analysis
how we look at data and summarize our findings
statistical inference
makes a statement about a population based on random/representative sample; includes a measure of how confident we are in our statement
experiments
researcher assigns subjects to certain experimental treatments
observational studies
researcher does nothing to subjects but observe x and y
anecdotal evidence
not good; untrustworthy source of data
census
official government data that can be easily accessible to the general public
samples
fast and cheap way to gather data/evidence of a specific population, but must be documented correctly so as to not imply that data collected may not be an accurate representation of an entire population
biased samples
BAD; volunteer sample (call in shows, internet polls), convenience sample (in class, libraries, bus stops)
good samples
random samples (subjects chosen by chance)
sample surveys
personal interview (good but costly), telephone (cheap but less effective), questionnaires (anonymous but low response rate)
margin of error
1 / n^2
1 / 1000^2 = 0.03 = 3%; too close to call b/c we are confident the true portion of people who will vote for Clinton is b/w 41% – 54%
A poll of 1000 randomly selected voters is conducted a week before the 2016 election. Results show that 51% of the sample is planning on voting for Clinton. Can you be confident Clinton will win?
true portion of people who will vote for Sanders is b/w 52% – 58%; we can be confident that Sanders will win
A poll of 1000 randomly selected voters show that 55% of participants plan on voting for Sanders. Can you be confident Sanders will win?
undercoverage
sampling frame is missing certain parts of population, leading to potentially inaccurate data results from a survey
nonresponse bias
those unwilling to participate in a survey could have different positions/opinions and complicate the data results
response bias
those willing to participate may give untruthful responses due to bad memory retention or simply lying
wording of questions
can influence participants by use of long/confusing questions or leading questions
response variable
variable that we measure and can draw conclusions from; basically what we are looking to prove in a given experiment
experimental units
the actual individuals (subjects) involved in the experiment
treatments
experimental conditions given to subjects; parameters of the experiments
comparative experiment
compare two or more groups to eliminate confounding and control variability of results
placebo
dummy treatment; psychological effects/treatments are important when dealing with experimental units
control group
group that receives placebo; helps determine true effect of treatment. not necessary when comparing more than one form of treatment
blind study
subjects unaware of treatment they are receiving
double blind study
both experimental units and people dealing with subjects don’t know what treatment each unit is being subjected to
random samples (randomization)
use a mechanical method to select subjects and assign them to treatments; can make use of probability to analyze results; avoids selection bias
replication
number of experimental units that get each treatment
cross-sectional studies
sample surveys that just want to take a “snapshot” of the population at current time
case-control studies
retrospective studies in which we match each case (POSITIVE OUTCOME) with a control (NEGATIVE OUTCOME) and then ask questions about the explanatory variable
prospective studies
forward thinking; follow subjects into the future
