What to know for the test on wednesday:Chapters 1-6• Individuals• Variable •Data• Know three types of studies:⁃ Observations - collect data but have not applied a treatment⁃ Sample Surveys⁃ Experiments - have applied some sort of treatment⁃ Know the differences and similarities⁃ Know what each of them are⁃ Know what a census is• Population versus sample⁃ Be able to identify the sample and the population from description⁃ Know the difference between the two• Poor Sampling Methods⁃ Convenience Sample⁃ Voluntary Response Sample - information is requested, people self select to participate in the survey⁃ Know that these two methods result in bias• Most Basic Good Sampling Method⁃ Simple Random Sample (SRS)⁃ Definition⁃ How to take• Parameter - measure that we're looking for⁃ generally will not know the value of a parameter⁃ attempt to estimate parameter using the statistic• Population• Statistic • Sample• Bias⁃ Know how to reduce bias⁃ Know definition• Variability⁃ Know how to reduce variability - bigger samples⁃ Know definition (how to spread out the values of the statistic are when we take many samples)•p hat⁃ how to compute ⁃ number of successes / total number of chances⁃ quick formula for MOE at 95% confidence for p hat⁃ 1/ square root of n⁃ know basics for confidence interval (estimate + or - margin of error (MOE))⁃ to get a smaller (narrower interval) -> larger sample size⁃ to get a smaller (narrower interval) -> lower level of confidence• Sampling errors - caused by the act of taking a sample⁃ random sampling error (MOE)⁃ bad sampling methods (voluntary response, convenience, etc)⁃ Incomplete sampling frame (know what a sampling frame is)⁃ Undercoverage (leaving out or not sufficiently taking into account part of a population)⁃ Undercoverage - left out part of the population⁃ doesn't take into account all parts of the population⁃ makes the sampling frame incomplete⁃ Sampling frame - a list of the population that we use for taking the sample• Non-Sampling Errors - errors not resulting from the act of sampling⁃ Processing Errors⁃ Response Errors (incorrect answers)⁃ Non-response Errors⁃ Response bias - answering the way they think the asker wants them to⁃ Poorly worded questions - lead someone to an inaccurate response• Stratified random samples ⁃ What it is and why we use it⁃ Identify groups in the sample - groups are called strata⁃ How to take it (stratify first, then random sample within strata)⁃ Identify strata, and within each strata, take a random sample• Cluster Sampling⁃ What it is and why we use it⁃ Identifying specific geographic areas, then we evaluate each subject within that area⁃ How to take it⁃ randomly sample cluster, then evaluate everything in cluster• Response and Explanatory Variables⁃ Know how to recognize and determine which is which⁃ Response - what was the effect⁃ how did the treatments affect the situation• Subjects (like individuals)⁃ the people that take part in the experiment• Treatment⁃ what is being done to subjects in an experiment⁃ Know that treatments are different levels and/or different combinations of explanatory variables• Lurking Variable⁃ Definition⁃ variables that are not part of our list of explanatory variables, but we are able to identify that they may have an impact on our study⁃ Know how we attempt to control for (account for) lurking variables from comparing two or more treatments• Confounded Variables• Placebo Effect⁃ takes place when somebody has received a treatment and they believe so firmly that there is gonna be an effect on them that they feel some kind of effect⁃ sugar pill⁃ pill that should not generate any kind of response• Randomized Comparative Experiments⁃ Compare two or more groups to control the effects of lurking variables⁃ Randomization (hopefully) controls for bias/lurking variables⁃ Large sample size control variability⁃ the larger the sample, the less variable• Statistical Significance⁃ Definition⁃ not likely to occur by chance⁃ you are never SURE⁃ How to interpret (like in homework)⁃ We can say there is a statistically significant difference when comparing groups by comparing confidence intervals⁃ Confidence intervals overlap -> NOT statistically significant⁃ Confidence intervals do NOT overlap -> statistically significant• How to live with observational studies⁃ some situations where you can not do experiments⁃ not ethically acceptable⁃ anything that causes harm to subjects• Double Blind - neither the subject nor the person evaluating the study is aware of the treatment received• Single Blind - one side or the other does not know• Problems with Real World Studies⁃ Undercoverage⁃ Refusal to Participate⁃ Non-adherers⁃ do not follow the rules⁃ can confound the outcomes of the study⁃ Dropouts⁃ people decide they don't want to do it anymore⁃ may not be able to stay in experiment• Goal is to generalize to the population⁃ Statistical Signficance⁃ Realistic Setting⁃ Repeatable Results• Randomized Comparative Experiments⁃ Compare two or more groups to control the effects of lurking variables⁃ Randomizing to control bias⁃ Using large sample sizes to control variability⁃ We studied three designs⁃ completely randomized design (analog to simple random sample)⁃ all subjects randomly allocated to a treatment⁃ Block design (analog to stratified random sample)⁃ helps reduce variability in estimates⁃ subjects divided into blocks based on similar characteristics⁃ controls for lurking variables⁃ Matched pairs⁃ paired samples⁃ special case where each pair is a block⁃ only used when there are two