Chapter 3 - elementary statistics textbook PDF

Preview

Summary

elementary statistics textbook...

Description

Chapter 3 Wednesday, January 16, 2019

• •

•

•

11:18 AM

Gaining information about the variables of interest Various designs for studying variables ○ Observational study- values of the variable or variables of interest are recorded as they occur naturally ○ Sample survey- individuals report variables values themselves, frequently by giving their opinions ○ Experiment- impose a treatment to manipulate values of the explanatory variable ○ Anecdotal evidence (avoid)- personal accounts Identifying study design ○ Retrospective- information about variables in the past ○ Prospective- information about variables values to occur in the future

Error in studies conclusions ○ Statistics is an inexact science ○ Consider the consequences of drawing the wrong conclusions in one way

•

•

•

•

•

•

or the other ○ Sample size plays a role in the type of error § Designing a study with a larger sample is a way to reduce the risk of concluding that "nothing is wrong" Sample surveys ○ Individuals report variables values themselves, frequently by giving their own opinions Source of bias in sample surveys ○ Open question- without a fixed set of response options ○ Closed question- provides or implies a fixed set of possible responses ○ Surveyors should always think carefully about whether questions should be open or closed- for closed questions great care should be taken to include all the reasonable options that are possible ○ Responses are more likely to reflect the truth if the survey provides respondents assurance that they are completely anonymous ○ Careful wording of questions can be helpful in eliciting honest responses about behaviors that people are usually reluctant to admit Observational studies ○ Confounding variables- one that clouds the issue of causation because its values are tied in with those of the so called explanatory variable, and also play a role in the so-called response variables values ○ In general, researchers control for the effects of a confounding variable by separately studying groups that are similar with respect to this variable ○ It is because of the existence of a virtually unlimited number of potential confounding variables that we can never be 100% certain of a claim of causation based on an observational study Paired or two-sample studies ○ Two-sample study design- compares responses for two independent groups ○ Paired study design- two response values recorded for each unit (matched pairs) ○ Help to control confounding variables by comparing responses for individuals occurring in pairs that are similar in important ways Experiments ○ The key to establishing causation is to ensure that individuals differ only with respect to their values of the explanatory variable ○ Treatment groups, control groups, subjects Randomized controlled experiments R h t l l f th l t i bl ith

Researchers control values of the explanatory variable with a randomization procedure ○ The groups tend not to differ significantly with respect to any potential confounding variable Double-blind experiments ○ Blind subject in an experiment is unaware of which treatment he or she is receiving ○ Placebo is an inert treatment given to duplicate the psychological effects of the actual treatment ○ Placebo effect- where subjects respond to the idea of treatment rather than to the treatment itself ○ Blind experimenter is one who when assessing the response of interest is not aware of which treatment a subject has received ○ Experimenter effect is the conscious or subconscious tendency of an experimenter to be influenced by knowledge of a subjects value of the explanatory variable when assessing the subjects response ○ Double-blind- both subjects and researchers are blind "blind" subjects" ○ Placebo effect "blind" experimenters ○ Recording its values is a straightforward process in which researchers need not use their own judgement in making an assessment ○ In general, the most reliable way to determine whether the explanatory variable is actually causing changes in the response variable is to carry out a randomized controlled double blind experiment Pitfalls in experimentation ○ Hawthorne effect- the phenomenon wherein peoples performance is improved simply due to their awareness of being observed ○ Lack of realism, noncompliance, treatments that are unethical or impossible Modifications to randomization ○ Blocking in an experiment entails first dividing the subjects into groups of individuals who are similar with respect to an outside variable that may be important in the relationship being studied ○ Can help ensure that the impacts of treatments are most effectively measured, without interference from background variables ○ Paired study to examine a relationship for a sample of blocks of just two individuals who are similar in many important respects ○ Before and after studies are a common type of paired design ○

•

• •

•

•

○

Before and after studies are a common type of paired design...