Study Guides (248,357)
Psychology (776)
PSYC 3000 (24)
all (3)
Midterm

# Test 3 Review.docx

15 Pages
67 Views

School
Department
Psychology
Course
PSYC 3000
Professor
All Professors
Semester
Winter

Description
Inferences about a Population Proportion: * Has the proportion in a group changed over time - Relationship; with confidence interval: approximate hypothesis test using interval; see if p Is in O the value - CI: Two sided: correspond to Two sided Tests Null: comes from context of problem; no effect P value: conditional probability, tells us probability of getting result, given, the null is true - If the p is low: the null must go - It’s the probability of the data; not the null 2 - Big P: Fail to reject the null Statically significant - > Set Alpha level: a: common levels: 0.01 0.05 0.1 (when two sided a/2) Alpha level = significance level “Reject the null at the 5% significant level” Hypothesis testing/confidence go hand in hand; - 95% confidence interval: Alpha 5% - Confidence interval 99%: A: 1% Conditions: - Independence assumption - Randomization condition - 10% Condition - Success Failure condition Making Errors - Type 1: The null hypothesis is true; we reject it (false positive)  Occurs at a level of a: 5% 10% 1% : unusual sample  Probability of type 1: a  Type II: the null is false; but we fail to reject it (guilty person free) False negative  Probability of Type II: B; decrease this by increasing a (but then you have more chance of type I) - Power: test’s ability to detect a false hypothesis  Jurors ability to properly convict  High power; we have looked deep enough  Power: 1 – B (type II) Effect size: the distance between the null: P anO the P - Ask ourselves how big of a difference would matter (flip a coin 50% maybe 60% would mean better than chance) - - Bottom has larger effect size: S tandardized mean difference between the two groups - cohens D: measure of effect size: diff mean sample – mean pop/SD - small D: 0.2 – 0.5 - Medium D: 0.5- 0.80 - Large 0.8 + We can lower alpha to lower chance of Type 1 error: 0.05 - > 0.01 - This increases B: chance for type II error - Therefore this lowers power - Bigger the effect size; more power; two models further apart; less overlap; less likely to make type II error - Increase sample size: Increasing sample size; increases power and decreases probability of Type 1 and type II  Tests with greater likely hood of Type I error: have more power less chance of Type II error Can increase: alpha -> get higher power; but also more chance of Type II Comparing Two Proportions (Two Proportion Z test) - We want to make inference between the difference two group proportions Model: Sampling distribution model for the difference between two proportions Sampling distribution modeled by - normal model - Mean:   p  1 2 - Standard Deviation: SD  1p ˆ2 p1 1 p2 2 n1 n 2 Assumptions and Conditions: - Indpendance assumption - Randomization - 10% condition - Independent Groups Assumption - Success/failure condition How do we test: - Test for no difference - H 0 p1– p 2 0 - What is this null value: P pooled - Then check CI see if it is a plausible value Inferences about means: Quantitate Means - one sample T-test (week 12 for example) - Based on: Student’s T model * if the parent population from which we are sampling is normally distributed: then the random variable has the student’s t distribution with (n-1) degrees of freedom Model: sampling distribution of the mean: centered at the population mean: u - law of large numbers; larger the samples size; closer we get to actual value: closer approaches normal distribution: will look more bell shaped (histogram) -  Now using Student’s T model: - t distribution: lower height; wider spread Objective: to determine whether the mean of sample differs from known/estimated population mean  Data Requirements: One sample/DV (quantative/numerical/Continuous) - As sample size increases the sampling distribution of sample means approaches normal distribution - CLT: sampling distribution of the mean is centered at population mean, μ Conditions: - Independence - Randomization - 10% Condition - Nearly Normal Condition: Make histogram/normal probability Plot  If data is normally distributed; data points will be close to the diagonal line - Large enough sample condition: (5-10) is enough for normal, if skewed minimum 30+ -> check the histogram -> Check the normal probability Plot -> Check Shapiro Wilk: If the P value of the Shapiro wilk test is greater than: 0.01, 0.05, 0.10: it is normal - if it is below these values: The data significantly deviate from normal distribution The Independent Sample’s T Test (week 12) - based on: sampling distribution for the difference between two means - We can first look visually: check out the box plots; see if means are different - See some difference; go to do inferential statistics - Based on Student’s T Model - Objective: Test whether the means of two groups differ significantly on some DV Requirements: - One IV: 2 separate groups - One quantitative DV - 0 not included in CI: reject the null * unlike proportions test; we don’t usually pool here Conditions: - Independent Group Assumption: Two groups must be independent one another - Nearly Normal Condition Check for both groups: Violation either one; violates  Check Separately - Independence assumption - Randomization Condition Parameter of interest: Difference between the two means Dependent Sample T test - Objective: Two test whether the means of two related groups differ significantly on some dependent variables Requirements: - One IV: Two related groups - One quantitative DV  just a one sample t test for the mean of the pairwise differences  once we know data is paired: we examine Pairwise differences  It’s the differences we care about; ignore the original 2 sets Conditions: - Paired Data assumption:  The data are paired - Nearly Normal condition:  Check histogram/normal probability plot of the differences Paired Samples And Blocks - > Scores in one sample related to scores in other sample: samples are dependent - Dependent samples: matched/paired samples - Example #1: Repeated Measures Design: Each Person tested more than once - means from each group of scores; from same sample (ex: focus in light and dark conditions) Example #2: Matched Subject Design - Matched subject design:  Separate groups: however subject in one group matched with an equivalent from another (match based on gpa; two groups; for IQ test)  Because again it is the difference we care about; we treat them as if they were the data Two sample problem becomes one sample problem d 0 tn1 SE d  sd SE d  n - Standard error for the mean applied to the differences Conditions mean: statistic follows: student’s t model: n-1 degrees of freedom Comparing Counts: Inferences About Categorical Data (Chapter 26) Goodness of fit - one group/ one categorical variable - is the observed distribution consistent with a theoretical model Test of Homogeneity - 2+ groups/One categorical variable - are the distributions of a variable the same in 2+ populations Test of independence: - One group/two categorical variables - Is there an association between the 2 categorical variables? Procedure: - Observed frequencies at each level similar to / different from the expected at each level  Observed: Actual frequency obtained for each level  Expected Frequencies: - Frequency values specified by the null hypothesis Frequency you would expect if null were true Chi Square Statistis: - Adding up the squares of the deviations between the observed and expected counts divided by the expected counts  it is a measure of discrepancy between the 2 distributions - if the two distributions are identical: chi square will be zero; the bigger the differe
More Less

Related notes for PSYC 3000
Me

OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Join to view

OR

By registering, I agree to the Terms and Privacy Policies
Just a few more details

So we can recommend you notes for your school.