Download presentation

Presentation is loading. Please wait.

Published byCody Turner Modified about 1 year ago

1
Eric Grodsky Sociology 360 Spring Lecture 12: Introduction to probability Review cross tabulations/ conditional distributions Conceptual introduction to parameters and statistics The idea of randomness revisited Thinking about probabilities Basic probability math

2
Eric Grodsky Sociology 360 Spring Race and affirmative action favor aff | action in | race hiring | white black other | Total str support | | 160 | | some support | | 112 | | oppose | | 402 | | str oppose | | 889 | | Total | | 1563 | |

3
Eric Grodsky Sociology 360 Spring Aid support by income quartile low-income | income category co | Q1 Q2 Q3 Q4 | Total def shd | | 539 | | prob shd | | 843 | | prob not | | 165 | | def not | | 56 | | Total | | 1603 | |

4
Eric Grodsky Sociology 360 Spring Parameters and statistics A parameter is the true population value for some attribute May be an attribute of a distribution (such as mean, median, variance) May be an attribute of a relationship (correlation, least squares regression line)

5
Eric Grodsky Sociology 360 Spring Assumptions When we talk about a parameter, we are assuming: The quantity exists in real life (empirical) The quantity is stable, at least for a moment The quantity is knowable These assumptions are not universally accepted

6
Eric Grodsky Sociology 360 Spring Measuring net worth The parameter of interest is mean net worth in the population of those over age 18 in the U.S. In 1990 There is an average net worth out there It is stable at a particular moment It is a knowable quantity

7
Eric Grodsky Sociology 360 Spring Voting for president Percent of “likely voters” supporting each candidate (October 11-13) Bush48% Gore44% Nader2% Buchanan1%

8
Eric Grodsky Sociology 360 Spring Assumptions on voting The parameters are percentage of the voting population voting for each candidate. The assumptions are: There is an actual percentage out there It is stable at a particular moment It is a knowable quantity

9
Eric Grodsky Sociology 360 Spring The internet stamp tax In a debate in the fall of 2000, Rick Lazio and Hillary Clinton were asked to share their views on a House bill to allow the U.S. Postal Service to tax at 5¢ a pop The bill is fictitious- an internet hoax Assume we asked Americans about their views on this bill

10
Eric Grodsky Sociology 360 Spring The internet tax question Recently, the U.S. House of Representatives took up a bill that would tax at 5¢ a message. Would you say you strongly oppose, somewhat oppose, somewhat favor, or strongly favor this legislation?

11
Eric Grodsky Sociology 360 Spring Assumptions on the internet tax The parameters are percentage of the population strongly opposed, somewhat opposed, somewhat in favor and strongly in favor of the bill. The assumptions are: There is an actual percentage out there It is stable at a particular moment It is a knowable quantity

12
Eric Grodsky Sociology 360 Spring Statistics and parameters A statistic is your best guess at the value of a parameter; your attempt to infer the parameter’s value We are almost always interested in parameters (properties of a population), but choose to estimate those parameters from samples This applies to both experimental and observational studies

13
Eric Grodsky Sociology 360 Spring Population and theoretical distributions There is a connection between population distributions and theoretical distributions Though we believe there are population distributions, we seldom if ever observe them Population and theoretical distribution share the same notation

14
Eric Grodsky Sociology 360 Spring Some (old) new notation It is important to distinguish between sample and population distributions So important we use different symbols MeanStd dev Samples Population

15
Eric Grodsky Sociology 360 Spring How are statistics and parameters related? If a statistic is calculated from data from a simple random sample, the distribution of the statistic has a known relationship to the population parameter. By building on chance, on probability, we can make claims concerning the parameters in which we are interested.

16
Eric Grodsky Sociology 360 Spring Randomness revisited To say something is random is to say that the outcome or value of something cannot be known with certainty before it is observed This is very different from saying that the distribution of this value is random Many phenomena are random at one level (individual observations), but follow a pattern at another level (aggregations)

17
Eric Grodsky Sociology 360 Spring The coin toss example Heads or tails is the outcome of each trial (flip of the coin) Assuming the coin is balanced, as the number of tosses increases, the proportion of heads approaches 0.50

18
Eric Grodsky Sociology 360 Spring The coin toss example

19
Eric Grodsky Sociology 360 Spring To find probability, you must: Have a long series of independent trials Observe and record the outcomes of those trials Aggregate across trials to find the probability

20
Eric Grodsky Sociology 360 Spring The voting example Think of each observation as a trial The responses are Bush, Gore, Nader or Buchanan Observations are independent if: probability of selection for each sample member is independent interviewer and response are independent

21
Eric Grodsky Sociology 360 Spring The voting example In simulated data, I gave each observation a 44% probability of choosing Gore. The following graphs plot the proportion of observations voting for Gore by the number of observations This is analogous to the number of successes (or failures) by the number of trials

22
Eric Grodsky Sociology 360 Spring Voting with 200 observations

23
Eric Grodsky Sociology 360 Spring Voting with 2000 observations

24
Eric Grodsky Sociology 360 Spring Probability math A probability model describes a random phenomenon, or a random event. Probability models begin with: The sample space (S), which is the set of all possible outcomes The event, which is any outcome or set of outcomes of interests A way of assigning probabilities

25
Eric Grodsky Sociology 360 Spring The sample space Think of a sample space as a population of outcomes. All possible outcomes are included. The number of possible outcomes varies with the number of trials

26
Eric Grodsky Sociology 360 Spring Examples of sample space: one observation Coin toss S={heads, tails} Voter poll S={Bush, Gore, Nader, Buchanan}

27
Eric Grodsky Sociology 360 Spring Examples of sample space: two observations, order counts Coin toss S={heads tails heads heads tails heads tails tails} Voter poll S={Bush BushNader Nader Bush Gore Nader Bush Bush NaderNader Gore Bush BuchananNader Buchanan Gore GoreBuchanan Buchanan Gore Bush Buchanan Bush Gore NaderBuchanan Gore Gore Buchanan Buchanan Nader}

28
Eric Grodsky Sociology 360 Spring Examples of sample space: two observations, order doesn’t count Coin toss S={heads tails heads heads tails tails} Voter poll S={Bush BushNader Nader Bush Gore Bush Nader Bush BuchananNader Buchanan Gore GoreBuchanan Buchanan Gore Nader Gore Buchanan }

29
Eric Grodsky Sociology 360 Spring Sample space, one outcome of interest, 10 trials Coin toss (number of heads) S={0,1,2,3,4,5,6,7,8,9,10} Voter poll (votes for Bush) S={0,1,2,3,4,5,6,7,8,9,10 }

30
Eric Grodsky Sociology 360 Spring The event An event is a sample of outcomes, a subset of interest to us. Not a random sample, a subset. If order counts, event might be “getting heads first, tails second” If order does not count, event might be “getting one heads and one tails” The sample space and event depend on the research question

31
Eric Grodsky Sociology 360 Spring Properties of probability Probability is the likelihood of some event occurring Any probability is between 0 and 1 0 P(A) 1 0 P(vote Nader) 1

32
Eric Grodsky Sociology 360 Spring Properties of probability The sum of probabilities for all possible outcomes is 1 P(S)=1 P(Nader or Bush or Gore or Buchanan or other)=1

33
Eric Grodsky Sociology 360 Spring Properties of probability The probability that an event does not occur is 1 minus the probability that an event does occur P(not A)=1-P(A) P(not Nader)=1-P(Nader)

34
Eric Grodsky Sociology 360 Spring Properties of probability It two events A and B have no outcomes in common, the probability of either event occurring is sum of probabilities of A and B. P(A or B)=P(A) + P(B) P(Nader or Buchanan)= P(Nader) + P(Buchanan) If the above is true, we call A and B disjoint events

35
Eric Grodsky Sociology 360 Spring Probabilities and frequencies Probability and frequency are closely related logically and mathematically The relative frequency for some event in a population is the probability of that event The relative frequency of some event in a sample is an estimate of the population probability

36
Eric Grodsky Sociology 360 Spring Who supports aid for college? low-income | race of respondent co | white black other | Total def shld | | 539 | | pr shld | | 843 | | pr not | | 165 | | def not | | 56 | | Total | | 1603 | |

37
Eric Grodsky Sociology 360 Spring Assigning probabilities Sometimes it is useful to assign probabilities rather than observing them Simulations (such as the voting analysis) Sampling When we assign probabilities, we often do so using random draws from a density curve

38
Eric Grodsky Sociology 360 Spring Uniform distribution (empirical)

39 Eric Grodsky Sociology 360 Spring Simulating voting In this case, 44% of “likely voters” in the Gallup poll preferred Gore, 48% preferred Bush Assign voters to Bush if P(X) .44 Assign voters to Gore if.44

40
Eric Grodsky Sociology 360 Spring Sampling voting Assign likely voters equal (uniform) probabilities of selection Select voters at random Observe their voting preferences In this case, 44% of “likely voters” in the Gallup poll preferred Gore, 48% preferred Bush

41
Eric Grodsky Sociology 360 Spring In both cases… Whether or not individual i prefers Gore, Bush, Nader or Buchanan is a random variable In the simulation, observations are randomly assigned to a candidate In the survey, observations have views and are randomly selected

42
Eric Grodsky Sociology 360 Spring Random variables “A random variable is a variable whose value is a numerical outcome of a random phenomenon” (Moore, p. 231) Not completely random Not necessarily mostly random Just needs a random component

43
Eric Grodsky Sociology 360 Spring Probability distributions The probability distribution of a random variable X shows us the values X can take There are many different probability distributions One with which you are familiar is the normal probability distribution, AKA “Table A”

44
Eric Grodsky Sociology 360 Spring Next time The normal probability distribution and The sampling distribution which is NOT an empirical distribution

45
Eric Grodsky Sociology 360 Spring Homework due Wednesday, March 28 Moore: 4.10,18,20,32,35,40,42,43,54

Similar presentations

© 2017 SlidePlayer.com Inc.

All rights reserved.

Ads by Google