Advanced Stats Unit 3: Collecting Data Study Guide

census

An attempt to collect data for the entire population

observational study

observes individuals and records variables of interest without trying to influence the response

survey

selects a sample to collect data. We use the sample to make inferences (conclusions) about the larger population of interest. A sample can be biased if poor sampling methods are used. 

sample

a subset of the population

biased

systematically favor one outcome over another

voluntary response sample

occurs when the respondents self-select to be in the survey. Usually only those with a strong opinion reply. This leads to biased results due to undercoverage

undercoverage

underrepresenting or excluding certain groups of a population

convenience sample

occurs when the individuals in the sample were the easiest ones to reach from the population. This often occurs when surveying individuals at a specific time and place, resulting in biased results because undercoverage occurs as a result of people at a specific time and place often sharing similar characteristics.

simple random sample (SRS)

consists of individuals from the population chosen in such a way that every individual has the same chance of being selected and every set of individuals has an equal chance to be the sample actually chosen. You can carry out an SRS by mixing names in a hat and drawing names. 

stratified random sample

divides the population into strata, groups of individuals that are similar on some characteristic that we suspect or know affects the response variable. Then a separate SRS is taken from each stratum. This ensures a more representative sample.

cluster random sample

divides the population into clusters, groups of individuals that are heterogeneous creating groups that are similar to each other with each one being similar to the population. Then an SRS of clusters is taken and all individuals in the chosen clusters are included in the sample. Cluster sampling can be more practical than other methods (faster, easier, cheaper, etc.)

systematic random sample

selects individuals at regular intervals, such as every kth individual from the population. Systematic random sampling can be easier than other methods but can also be biased if there are patterns in the way the population is ordered that can coincide with the regular intervals.

sampling frame

a list of the individuals from which the sample is chosen

sampling error

occurs (even in a well designed study) because a sample is not perfectly representative of a population, due to random chance in selection. Sampling error can be reduced by increasing sample size. 

nonsampling error

occurs because of nonresponse, response bias, poor wording of a question, interviewer bias, and poor survey design (undercoverage, OR using a poor sampling method like voluntary response or convenience sampling)

nonresponse

when people you want in your survey don’t respond

treatment

An experiment has a __________, a condition to be imposed on the subjects. A treatment is what distinguishes an observational study from an experiment.

confounded

Variables can be ____________ when you cannot separate the effects of the variables from each other on the response variable. We want to take steps to avoid confounding.

lurking variable

A variable that is not among the explanatory variables that may influence the response variable. We want to avoid lurking variables being confounded with the explanatory variable. 

Experimental Units

The smallest collection of individuals (not necessarily people) to which the treatments are applied. If the experimental units are people, they are also called subjects.

treatment

an experiment in some condition being imposed on the experimental units. Each _______ can have a number of factors (explanatory variables). Each factor can have different levels. The number of treatments is determined by the number of combinations of the different levels of various factors.

levels

the different values or options it can take on (temperature might be a factor/explanatory variable with three levels: low, medium, high)

double-blind

A study is _______ if neither the patient nor those interacting with the patient know which treatment the patient is receiving.

single-blind

A study is _________ if the patient or those interacting with the patient don’t know which treatment they’re receiving

Basic principals of experimental design

comparison, control, random assignment, and replication

comparison

To use _________ by contrasting two or more treatment groups. One of those treatment groups may be a control group which serves as a baseline for comparing the effects of the other treatments.

control

To ________ for lurking variables that might affect the response by keeping their values of constant for all treatment groups. This helps to ensure that the only systematic difference between groups is the treatment being given.

random assignment

To use _________ by randomly giving out the treatments to experimental units to create roughly equivalent treatment groups by balancing the effects of any remaining lurking variables that cannot directly be controlled. 

replication

To use __________ by using enough experimental units in each treatment group so that any differences in the treatment groups are more likely to be due to the treatments and not random chance.

placebo

A fake pill or treatment. Often the control groups receives a placebo

Placebo Effect

Occurs when patients have a real response because they believe they are receiving a treatment even though the treatment they received has not active ingredient. 

randomized block design

To control for a lurking variable by creating groups, or blocks, of experimental units that are known to be similar on some characteristic that are believed to affect the response. A block design (an experimental design) is analogous to a stratified random sample (a survey design)

matched pairs design

a special case of block design in which you create blocks by matching pairs of experimental units on many characteristics and randomly assign the treatment within each pair. 

inference

the conclusions we can make from a study

statistically significant

results of a study are said to be _________when an observed effect is so large that it would rarely occur by chance

scope of inference

the ________ of a study refers to the extent to which the conclusions of a study can be generalized

-if random sampling is used in a study, we can generalize the results of the study to the larger population from which we sampled

-if a treatment was randomly assigned in a study, then we can make cause-and-effect conclusions

generalized

who or what we can draw conclusions about and whether we can infer causation

sampling variability

the idea that different random samples yield different results. We can reduce sampling variability by increasing the sample size.