SOC SCI 3A Study Guide - Midterm Guide: Level Of Measurement, Correlation Does Not Imply Causation, Response Bias
Lecture - 2/5/18
What causes problems within an experiment?
1. Bias: information was misleading, consistently led into the same wrong direction
2. Chance error: one thing results in one outcome and results in a different outcome
another time; things just occur by chance
3. People not knowing what information was saying: read the information wrong,
making wrong assumptions → logic doesn’t follow (conclusion doesn’t follow)
Bias:
● The answer is not true/is wrong
● We need to acknowledge that it’s wrong
Ex. Article: Yale University, 20 years after they graduated (looking at early 40s men)
● Survey: how much money are you making?
● Average responses: $25,111 (worth $251,110 today) → the number is not true; it does
not reflect reality → bias
● Model:
○ 1. Lie
○ 2. Sampling the top half (not a random sample)
○ 3. Wealthy have telephones: done by calling → it’s easier to find the wealthy
Types:
●Response bias: deals with anything that affects your answers (ex. Wording, order of
questions, gender and ethnicity, times, the ways of asking)
a. Presence of lying (ex. Driver’s License: actually 173 pounds, but most likely to
say 170 pounds)
●Non-response bias: people get the survey, but do not respond; people who don’t
respond are different from the people who do (maybe the people who make $30,000
respond and those who make $2,500 do not)
●Selection bias: was able to contact people that are easier to find → it was hard to find
people from 70 years ago
a. Ex. People who join as alumni are typically rich people (e.g. mayor of the town,
doing something famous, etc. → easier to find)
2 types of measurements
1. Median
2. Mean
a. Mean affected by extreme outliers, median is not
b. Sometimes median is more important, sometimes the mean is more important →
listen to the information → Are you being misled?
Q: In OC, the average house price is the median
Document Summary
We need to acknowledge that it"s wrong. Article: yale university, 20 years after they graduated (looking at early 40s men) Average responses: ,111 (worth ,110 today) the number is not true; it does not reflect reality bias. Sampling the top half (not a random sample) Wealthy have telephones: done by calling it"s easier to find the wealthy. Response bias: deals with anything that affects your answers (ex. Wording, order of: presence of lying (ex. Driver"s license: actually 173 pounds, but most likely to questions, gender and ethnicity, times, the ways of asking) say 170 pounds) Non-response bias: people get the survey, but do not respond; people who don"t respond are different from the people who do (maybe the people who make ,000 respond and those who make ,500 do not) Selection bias: was able to contact people that are easier to find it was hard to find: ex.
