Causes of Error in Sampling Sampling Error Sampling
- Slides: 19
Causes of Error in Sampling
Sampling Error • Sampling error is error caused by the way you chose your sample – Volunteer Sampling & Convenience Sampling – Causes Bias • How do you reduce? – Use randomness in choosing sample your
Random Sampling Error • Random Sampling Error – You sample may randomly have a higher or lower percentage of females, college educated people, Hispanics, etc. than what is found in your population. – Causes Variation • How do you reduce? – Choose a larger sample size.
Random Sampling Error • If your sample ≤ 10% of the population than your error is about n = sample size – True no matter how big the population is • If you sampled 10, 000 Indiana residents, population 6. 5 million, what is your error? • If you sampled 10, 000 South Bend residents, population 101, 658 what is your error?
Random Sampling Error • If your sample ≤ 10% of the population than your error is about n = sample size • If your sample ≥ 10%, then your error is better than this equation, but we won’t get into that until chapter 9.
Confidence Interval • If you calculate your error to be 1%, your actual error could be larger. That 1% is what we call a confidence interval. You are 95% confident that the error between your measured result and the actual result is 1%. • If the results from your study show that the percent of Americans that believe in Heaven is 88 + 5%, what that means is you are 95% confident that between 83% and 93% of the population believes in Heaven.
Undercoverage • Undercoverage is when some groups in the population are left out of the process of choosing the sample. • What groups CAN’T be contacted by a survey where you call random people and ask their opinion? – Amish – Homeless – People serving overseas – Prison Inmates – Unlisted numbers & other people without phones
Undercoverage • Undercoverage is when some groups in the population are left out of the process of choosing the sample. • How do you reduce/prevent undercoverage – Census (poll everyone) – Reduce bias in your sampling (avoid convenience & volunteer sampling) – Stratified Sample: Split the population in groups and then sample each group. • Example: Instead of randomly selection 100 people, randomly select 50 men and 50 women
Nonsampling Error • Processing Error – Enter a number wrong in Excel – Math Error • How do you reduce/prevent – Double Check your Work – Don’t Rush
Nonsampling Error • Response Error – given an incorrect response • Why would someone do that? – Lie • How much do you weigh? • Have you ever used drugs? – Remember Incorrectly • How many minutes have you watched TV this week? • Where exactly were you at 3: 30 PM last Saturday? – Vague or Confusing Question • How many windows do you have? Do door windows count?
Response Error • Lying causes bias • How can we reduce lying? – Confidential: the interviewer promises their name won’t be released with the results – Anonymous: even the interviewer doesn’t know which response sheet corresponds to which person – Study them without them knowing, as with the handwashing homework problem • Is this ethical?
Response Error • How can we reduce memory errors?
Response Error • Confusing questions cause bias if they favor one answer over another. They cause variability if they are just confusing in general, but don’t favor a particular outcome • How can we reduce people from misunderstanding the question? – Clear, careful and extremely specific wording
Response Error: Question Wording • “Owning stock” means different things to a Wall Street Banker & a Texan Rancher • 13 % of Americans think we are spending too much on “assistance to the poor, ” but 44% think we are spending too much on “welfare” • A poll in Scotland showed that 51% would vote in favor for “independence for Scotland, ” but 34% would vote in favor for “an independent Scotland separate from the United Kingdom. ” • Assistance & Independence are positive words while Welfare & Separate are negative words
Wording Questions • Loaded questions cause bias – Do you favor banning private ownership of handguns in order to reduce the rate of violent crime? – George Bush: great president or greatest president? – Do you support our president? – Do you agree with all of Obama’s policies? – Do you approve or disapprove of the way Barack Obama is handling his job as president?
Wording Questions • Open vs. Closed Questions – Rate Obama’s performance on a scale of 1 to 10 vs. what do you think about Obama’s job as president? – Limiting people’s options can cause bias – Limiting people’s options can reduce people misunderstanding the question – Closed Questions are easier to analyze
Non Response • They refuse to answer your survey. They hang up on you, slam the door in your face or just politely say no.
Nonadherer • A nonadherer is a generic term for someone who doesn’t follow directions for whatever reason. – Example: They forgot to take the experimental pill every 6 hours. – Example: They lied on their survey form • How would you reduce?
Dropouts • Sometimes your study the same person over several days, weeks, months or even years. Dropouts are people who start doing the experiment, but then stop half way through.