Session 4 - How to Analyze Rates and Proportions.docx

6 Pages
110 Views
Unlock Document

Department
Cell and Systems Biology
Course
CSB345H1
Professor
William Navarre
Semester
Fall

Description
HMB325H © Lisa| Page 113 S E S S I O N 4 : H O W T O A N A LY Z E R AT E S A N D P R O P O R T I O N S ( C H A P T E R 5 ) LEARNING OBJECTIVES 1. recognize & describe the characteristics of the types of data (i.e. levels of measurement) associated w proportions 2. describe & calculate descriptive statistics used for proportions, including the standard deviation, standard error of a proportion, & confidence interval 3. describe & complete a difference of proportions analysis, including a confidence interval 4. describe Yates correction for continuity & its rationale 5. describe & carry out contingency table analysis 6. describe & determine the chi-square (χ ) test statistic 7. describe & carry out the Fischer exact test RATES & PROPORTIONS 1. information not measured quantitatively 2. measured categorically (ex. nominal or dichotomous scales) ex. male/female, alive/dead 3. no arithmetic relationship bw categories 4. need ways to describe/summarize & analyze this info DESCRIPTIVE STATISTICS 1. summarize by counts or proportions (percentages) ¿withfactor m ‘average’()= ¿∈wholesample = n 2. measure data’s variability 1. average ‘distance’ from the ‘mean’: −¿ mean ' 1¿ ¿ ' ('mean )¿ √¿ 5. ‘variability’: 'S D = (proportion)(1−proportion) √ 6. proportions range from 0% (no one has it) to 100% (everyone has it) 1. at what proportion is variability lowest? when is it highest? 1. lowest – everyone in the sample has the exact same results (no variability), at the extremes 1. 0% – no one has the outcome of interest 2. 100% – everyone has the outcome of interest 3. SD = 0 2. highest – in the middle, 50% STANDARD ERROR OF PROPORTION 7. measures how well a sample proportion estimates the ‘true’ (population) proportion (the variability) among diff sample proportions (aka ‘percentages’) SD SE= 8. recall, √ n 9. so, HMB325H © L| Page 2 2013 √(percent) 1−percent) (percent 1−percent) SEP= = n √ n percent = proportion 10. SEP enables calculation of confidence interval of a proportion (using normal distribution) HMB325H © Lis| Page 3013 11. ex. CI of a proportion: A survey of 106 patients indicates that 32 (30.2%) are still experiencing pain a month after their surgery 1. 95% confidence interval: SEP: (p 1−p ) (.302 .698 ) = =.045(4.5) √ n √ 106 CI: 95 CI=p±1.96(SEP)=.302±1.96(.045)=.215¿.389 = 21.5% to 38.9% 1. just over 30% are still experiencing pain, & the CI tells us that the true percentage of patients experiencing pain will range from as low as 21.5% to as high as 38.9% w an accuracy of 19/20 (95%) DIFFERENCE OF PROPORTIONS 2. ex. death after halothane vs morphine: does the risk of post-operative death differ? 1. is this difference consistent w a real clinical effect or simply random variation 2. need a statistical test enabling us to measure the probability of a difference this extreme or more extreme arising by chance t= differenceof means 3. recall the t test assessed the difference bw 2 groups: SEof difference 4. in the case of proportions, this would be: (p −p ) 1 2 (SE p1+SEp )2 5. plus 2 ‘refinements’ & 1 assumption: 1. estimating & using a ‘pooled proportion’ pooled– rather than using the 2 proportions separately 2. Yates correction for continuity – dealing w continuous data in t tests & ANOVA, dealing w simple counts (step-wise increments) in proportions 3. ‘large sample’ assumption ESTIMATING & USING p pooled 6. if the hypothesis that the 2 samples are drawn from the same population, then each sample provides an independent estimate of that pop proportion 7. as a result, we can estimate a single, pooled estimate of the proportion as: (m1+m 2 p pooled (n1+n 2 ‘LARGE SAMPLE’ASSUMPTION (p −p ) 1 2 8. because the ratio (SE p 1SEp ) 2 is approx normally distributed only for sample¿¿>5 ppooledample¿5 sufficiently large sample sizes, and (1−p pooled¿ ppooledroportion HMB325H © Li| Page 42013 9. the Fishe
More Less

Related notes for CSB345H1

Log In


OR

Join OneClass

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

Sign up

Join to view


OR

By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.


Submit