Flashcards
Exam 1 Part 3
Cards in this deck (16)
- 1Causation• Absolute causation
◦ Every time x happens then y
• Probabilistic causation - focus in political science
◦Can’t prove
◦The more money you make, the more likely you will be republican
◦Looks at probability and likely hood
◦What helps us predict a dependent variable
◦Never exact - 2Classic experimental design• We create an experiment where we control the conditions in which the variables interact
• You divide subjects into two groups - experimental group and the control group
◦Give them the same pre test and then provide the stimulate to the experimental group and
give them a post test
◦Then to find the effects
‣ You do the effect= (post testE - the pretextE)- (post testC - pretestC) - 3Solomon two control group research designWant to make sure that the pre test isn’t influencing the results so you are hoping that if you
compare those who had the stimulus are the same. Then you know the pre test isn’t influencing
your results - 4Solomon three control group research designThe third control group only gets the post test, If they aren’t the same it shows that there is an external factor influencing your experiments
- 5Precision matching◦Matching subject for subject. One person in your control group matches closely to someone
in your experimental group
◦Need a large enough group
◦You need an 65 year old man who is a contractor in both groups
◦As you get more control groups, you need more people to match
◦Rarely used since it’s hard, there is a struggle with what the external factors may be that you didn’t notice - 6Frequency distribution◦Each group has the same average characteristics
◦Equal number of male and female or young and old
◦Some disadvantages as it’s not super precise but general, there is a struggle with what the external factors may be that you didn’t notice - 7Randomization◦Safest way to make sure that the groups are similar
◦Much easier to assign
◦Likely to be representative - 8Field experimentsnot done in a lab but are done in the natural environment. Not in isolation
- 9Quasi-experimental designnot able to meet the criteria for a experimental design, you have an
experimental group but not a true control group so you try to stimulate one that would act like a
control group
Examples:
Post facto experiment
‣ Creating a research design after the fact
‣ Interested in seeing if people with a college degree are more likely to vote then those
without a college degree. We look at people that went to college vs people who didn’t
• Not a true experimental design as there may be factors that dictate why someone
may or may not go to college
◦Time series design
‣ Where the control and the experimental are one and the same at different times
‣ Same people before and after - 10Miller and Robyn article-the purpose of the research
◦If mail would cause people to vote more favorable for one candidate
the research question-
◦Does sending out more info about a candidate create more support for the candidate
(More info in notes day 4) - 11Scaling techniques◦We create scales
◦Combining a number of indicators into a single summary measure - 12Likert scaling◦Summary indicator of a trait based on respondents answer to the degree which they agree or
disagree to a series of statements
‣ Here is a question, do you strongly agree, agree, neutral, disagree or strongly disagree
‣ Using this we then add it up and divide
◦This scale is an over simplification as it’s very limited and it’s assuming that very single
question is equally important
◦Using ordinal level data but acting as if it’s interval level data
‣ It’s acting as if there are standard units between strongly agree and agree but we don’t
know that. We don’t know how far the variables are for the people but we can’t like we
do. So it’s ordinal data - 13Guttman scale◦Overcome some of the problems of the likert scale
◦Attitudes or behaviors are related
◦Uses a series of statements you either agree or disagree with (2 options)
◦If you hold a position, it should influence the opinion you hold on the next position
‣ Do you like the idea of using halfway houses to rehabilitate criminals?
• Agree or disagree
‣ Would you mind if there was a halfway house in your community
• Agree or disagree
‣ So the questions get more specific and difficult, we can order the level of intensity to
find those who strongly agree, agree, etc.
◦The problem is that while most people would fit on your scale but some people won’t fit that
and have answers that seem to contradict
◦So we have to make a guttman coefficient of reproducibility - 14guttman coefficient of reproducibility◦We try to find out where to put the outliers in our model
‣ We do that by finding the min amount of changes we need to make to fit them into our
score
‣ Then we use the Guttman coefficient of reproducibility to find out I’d we made to many
changes that we should just throw away the result
‣ If it’s .9 or higher than you can keep your scale, lower than it you should get rid of your
scale
Cr = 1- (the sum(number of errors)/number of people*number of questions) - 15Standardizing indicators◦Uses to develop a scale if your indicators do not have common units of measurements
◦Set each indicator to a 100 point scale
‣ This weights the indicators equally
‣ Set the highest case to 100
‣ Divide 100 by the value for the highest case
‣ Multiply all other cases by the number produced in step 2
‣ You then add up all the standardized scores and divide them by the number of
indicators which gives us the scale
◦When you standardize a score, you no longer have a number that is represented by
something like dollars or years, it doesn’t have a specific meaning
◦An issue with this is that it makes the categories all equally weighted and excludes other
factors
‣ You can weigh the differently - 16—
Information
- Description
- Government Course Political Research (Exam 1) Flashcards on Exam 1 Part 3, created by Talia Vinson on 03/10/2022.
- Created by
- Talia Vinson
- Views
- 0
- Rating