CH.4 - Gathering Data

Questions: How to gather data? How to ensure the data collected lets us make valuable conclusions later?

What are the 2 basic ways to collect data?

1. Conduct an experiment

2. Take an observational study

Experiments do what?

• Attempt to manipulate or influence the subjects to obtain data

• Researchers control what is given to subjects

• Subjects are randomly assigned groups

A properly designed experiment(s) can be used to ___

prove causation

What are observational studies?

Measure characteristics of the subjects without attempting to manipulate or influence the subjects

• cannot be used to prove causation

• concludes 2 variables are related

Difference(s) between experiments and observational studies

Experimental studies = subjects are assigned to a group and to its treatment

Observational studies = no subject is assigned to a particular group or treatment; subjects are doing their own thing

Which study can prove that the explanatory (independent) variable has an effect on the response (dependent) variable?

experiments

Goal of observational studies

describing the sample shown and generalize the characteristics of the sample to the larger population of individuals

To rely on the sample to tell us about the population, what must the sample be?

The sample must be representative of the population

Simple Random Sampling

• Each subject in the population has an equal chance at being in the sample

• Randomly selected

• Most basic and unbiased

• Doesn’t overestimate or underestimate the parameter

Stratified Sampling

• Population is divided into non-overlapping groups

• Simple random sampling is applied to each group

• Taking a few subjects from all groups

Cluster Sampling

• Population is divided into non-overlapping groups

• All individuals within a randomly selected group or groups are sampled

• Taking all subjects from a few groups

Convenince Sampling

easiest to obtain info

Systematically Sampling

using a rule to select a sample

What is bias?

Sampling methods that consistently produces samples that inaccurately estimates the parameter; a property of a sampling method, not data itself

Nonrandom sampling includes bias, what are the sample types involved?

Convenience samples and Voluntary response samples

Convenience samples

contains bias; subjects are easily available; restrictive with no randomize or adequate representation

Voluntary response samples

contains bias; subjects elected themselves into the sample; unrestrictive; mainly extreme responses; before a survey

Undercoverage

Certain demographics in the population are excluded in the sample

Nonresponse bias

subjects who cannot be reached or refuse to participate in the sample

Response bias

subjects don’t answer truthfully or questions are confusing/ misleading; during a survey

A larger sample doesn’t reduce or removes bias it will ___

create more bias because the process is already flawed

Experimental unit is another name for

subjects (people or objects that get the treatment)

What is a treatment?

condition applied to the subject

What is a explanatory (independent) variable?

influences changes in the response variable

What is a response (dependent) variable?

variable of interest, measured after treatments are applied

What is expected from experiments?

To have controlled studies with treatments being randomly applied to experimental units or subjects

Goal of an experiment

determine if the explanatory variable effects the response variable

Placebo

“Dummy” treatment; creates a baseline

Which group receives the placebo

control group

Double blind

Both the researcher and subject know who’s getting the treatment

Single blind

Researcher knows which treat is being given to the subject but subject doesn’t know

Purpose of random assignment is to ___

even out any lurking variables

Lurking variable

an unforeseen variable; variables relating to both the explanatory and response variable

Completely Randomized Design

Subjects are randomly assigned to treatments without concern for anything else

Match-Paris Design

Subjects are related or matched before the experiment takes place; either half of the pairs receives one of thte two treatments or two halves receives the treatments in different orders

Before - After Analyzation

• measure response variable before treatment is applied, measure again with the treatment after and then pair them in a matched-pair design

• Same number of observations is required

What is the conclusion when a randomized experiment finds an effect?

There’s a difference in treatments that caused the effect

Is association evidence possible of causation?

Yes, however, it isn’t enough to prove cause-and-effect

Random assignment

• performing experiments

• Benefit: concludes causation between treatment groups

• In theory, random assignment evenly distributes lurking variables among treatment groups, making the difference between treatment groups the treatment they receive

Random selection

• collecting subjects for a study - sampling from a population

• Samples are useful as long as they closely represent the population

• Generalization of the sample results to the population

Random assignment is used to

determine which observational (experimental) units go into which explanatory variable groups

Goals for random assignment

• Produce groups similar in all aspects except for the explanatory variable

• If groups differ significantly on the response variable, that can be from a cause-and-effect relationship between the explanatory and response variable

Describe the difference in randomness between random sampling and random assignment

Random sampling = selecting a representative sample from the population and making a generalization on the population from the sample

Random assignment = using similar treatment groups, a significant difference in the response variable can come form the explanatory variable; inferring causation