Download presentation
Presentation is loading. Please wait.
1
Stat 217 – Day 2 Selecting Samples (3.2)
2
Announcements Complete survey Computer usage rules
Studio open hours TR 10am-12, W 6-8pm Questions on syllabus? warning about coursenotes Advice for doing well in course 8-12 hours/week “participate in class,” “stay on top of material,” “ask questions”
3
Last Time Examining numerical information with a critical eye!
4
Last Time Collecting your own data
1. Defining variables/observational units characteristic that changes from individual to individual quantitative vs. categorical 2. Deciding on the measurement tools, definitions 3. Choosing the individuals population vs. sample
5
Example 1-5: Sampling CP Students
Sampling frame registration list addresses from Cal Poly website (b) Variables GPA (what is your GPA? average GPA?) major (what is your major? most population major?) whether or not the student smokes (vs. percentage that smoke)
6
Example 1-6: Chocolate Lovers
(a) lifespan = quantitative chocolate eating habits = categorical? smoking habits = categorical? weight = quantitative (b) 7,841 Harvard male graduates (c) men (d) bias? (e) cause and effect?
7
BIAS The sampling design is biased if sample tends to overrepresent certain segments of the population Justify why you believe the sample results will lead to an overestimate or an underestimate of the actual value for the population.
8
BIAS Common sources convenience samples voluntary response nonresponse bad “sampling frame” If we want our result to “represent” the population, have to be very careful when we try to select our sample
9
Today How do we select our subset so we believe the sample is representative of the population? “representative” = has the same characteristics
10
Example 2-1: Sampling Words
Do you think the lengths in your sample of 10 words is representative of the lengths in the population? How decide?
11
Example 2-1: Sampling Words
Population = all 268 words in the Gettysburg Address Sample = 10 words you selected Compare the proportion of long words in your samples to the population proportion (f) 99 long words population proportion = 99/268 =.369
12
Class Results “Dotplot” = each dot represents one of your samples
Bias = systematic tendency to choose longer words, overestimated the proportion of long words population proportion (i) Why?
13
Better Ways to Select Sample
Need list of every member of population “sampling frame” Example: telephone books, vehicle registration lists not a good choice in 1930s Example: registrar’s list 2. Number every member of the list statweb.calpoly.edu/chance/stat217/address.html
14
Better Ways to Select Sample
Table B Read off 3 digit numbers until you have 5 Skip numbers > 268 Eliminate any repeat numbers Report the corresponding “individuals”
15
Better Ways to Select Sample
Combine results To examine long-run properties statweb.calpoly.edu/chance/applets/applets.html
16
Behavior of Samples If randomly select the samples, results will not be biased Though results vary from sample to sample, are not consistently off in the same direction Results from larger samples are closer together Increases precision, does not decrease bias Population size does not matter!
17
Random Sampling Error Is variability from sample to sample around the population value will be able to measure the amount of random sampling error example 2-3 sample size is what matters, not population size or proportion of population sampled Example: Tasting soup
18
Does this solve all of our problems?
19
Does this solve all of our problems?
Nonsampling errors = select a random sample but then… missing data response errors lying nonvoters undecided response bias (sensitive question) wording of the question processing errors
20
Overcoming nonsampling errors
Repeat visits Insure confidentiality Test different wordings Change order of choices Give background material/encourage respondents to admit they don’t know Control appearance of interviewer
21
Lab 1 Start analyzing data population sample descriptive statistics
inference evaluate some conjecture about the population
22
Lab 1 Variable: choose Coke or Pepsi?
Summary: how many students choose Coke? Is this convincing evidence that a majority of students prefer one soda over the other?
23
Assignments By Friday By Monday Complete Blackboard survey
Look through Lab 1 and bring floppy disk Finish HW 1
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.