Section outline

  • Collapse all Expand all

  • Monday, April 3, 2017; 10:30 AM - 12:30 PM

    Lecture: Causal Inference in the Context of Observational Data: identifying threats to validity

    This class introduces the overall framework of causal inference from observational data and compares the motivation typically given in modern epidemiology with traditional accounts of causation, including the very influential Cook & Campbell framework and the traditional Doll & Hill criteria. 
     

    Faculty:  Maria Glymour

    Location:      Mission Hall 1406

    • Session Slides:

    • Lecture 1 Slides File

    • Session Audio/Video Recording (Access restricted to registered students):

    • Watch URL

    • Required Reading

       Winship C, Morgan SL. The estimation of causal effects from observational data. Annual Review of Sociology. 1999;25:659-706.

      ONLY NEED TO READ PAGES 659-669: This is an excellent paper but takes a lot of work to get through. 

    • Cook T, Shadish W, Wong V. Three conditions under which experiments and observational studies produce comparable causal estimates: New findings from within-study comparisons. Journal of Policy Analysis and Management. 2008;27(4):724-750. File
      Not available unless: Your ID number contains 02
    • Winship C, Morgan SL. The estimation of causal effects from observational data. Annual Review of Sociology. 1999;25:659-706. File
      Not available unless: Your ID number contains 02

    • Optional Reading:

        Cook T, Campbell D, Shadish W. Experimental and quasi-experimental designs for generalized causal inference: Houghton Mifflin; 2002. chapter 2, pg 37-

         

         

      1. Assignment: Optional narrative description

         

    • Monday, April 10, 2017; 10:30 AM - 12:30 PM

      Lecture:  Clustered Data Arising from Repeated Measures or Contextual Effects
       This lecture will discuss using random effects/multilevel models for neighborhood effects estimation.

      Faculty:  Maria Glymour

      Location:        Helen Diller Family Cancer Research Building HD-160

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Singer, J. D. (1998). "Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models." Journal of Educational and Behavioral Statistics 24(4): 323-355.
        2. Arcaya M (2013) "Effects of Proximate Foreclosed Properties on Individuals’ Weight Gain in Massachusetts, 1987–2008"  Am J Public Health
        3. Lee (2012) "Length of Inpatient Stay of Persons With Serious Mental Illness: Effects of Hospital and Regional Characteristics" Psychiatric Services 63, pg 899.
        4. Greenland S (2000) Principles of multilevel modeling. Intl J Epidemiology. 29: 158.

        Singer et al is a classic and brilliant article by one of the great popularizers of multilevel models.  It is worth reading several times.  Arcaya and Lee are examples of common applications of multilevel models to illustrate the types of questions people approach with these models.  Lee in particular illustrates how the lowest unit of observation does not need to be an individual. 

        Greenland's framing is unusual but extremely helpful because it makes the link between multilevel models and Bayesian frameworks. 

      • Arcaya File
        Not available unless: Your ID number contains 02
      • Lee File
        Not available unless: Your ID number contains 02
      • Singer File
        Not available unless: Your ID number contains 02
      • Greenland File
        Not available unless: Your ID number contains 02

      • Optional Reading:

      • Assignment: Find any article using clustered data and describe: the unit of clustering; the hypothesized effects and the level at which the exposure is measured (is it a characteristic of the cluster or the observation within the cluster); and the statistical model used to estimate the effect.  Describe whether there are any other statistical models that might be appropriate and whether they would be preferable (e.g., GEE vs mixed).

        Additional Assignment for Data Application IS: 

        1) Register to use data at: https://usa.ipums.org/usa/.  

        2) Download a data set from the 2000 Census 5% sample, including at a minimum basic demographics, year of birth, education, state (FIPS), and all available disability variables.  Feel free to include anything else you'd like. 

        3) Open the data set in the statistical software of your choice.  I suggest that you write your code to pull a 1% sample of the data, so you can manipulate it.  There are many ways to complete this selection, for example generate a random number with a uniform distribution and keep only those observation with random number <.01.  You may find that you cannot open the data set at all on your computer.  If so, go back to the IPUMS data set and select "customize sample sizes" and change the requested density.  

        4) Create a variable that is the % of people with <=6 years of education in each state (state low education).  

        5) Estimate a linear regression (not accounting for clustering by state) using education and state low education predicting self-care disability. 

        6) Estimate a mixed model with random intercepts for state and no fixed effects.  

        7)  Estimate a mixed model with random intercepts for state and own education as a fixed effect.  

        8)  Estimate a mixed model with random intercepts for state, own education as a fixed effect, and state low education as a fixed effect.

        9)  Next week we will ask you to write a summary of what you have done.  For this week, just try it and figure out all of the problems you will have getting the data, handling the data, specifying and interpreting the models. 

      • Quiz for Lecture 2 File
      • Lecture 2 Slides for 2017 File

    • Monday, April 17, 2017; 10:30 AM - 12:30 PM

      Lecture:  Longitudinal study designs: data sources

       The goal of this lecture is to consider how alternative data sources have different strengths and weaknesses that make them (in)appropriate for a research question. It is also to familiarize you with some major categories of study design.

      Faculty:  Maria Glymour

      Location:        Mission Hall 1406

      • Session Slides:

      • slides for lecture 3 File

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Wu T. The Atherosclerosis Risk in Communities (ARIC) Study: Design and Objectives. American Journal of Epidemiology. 1989;129(4):687.
        2. Pagidipati and Gaziano. Estimating deaths from cardiovascular disease: a review of global methodologies of mortality measurement.  Circulation. 2013; 127:749.
        3. Pickett KE, Luo Y, Lauderdale DS. Widening social inequalities in risk for sudden infant death syndrome. American Journal of Public Health 2005;95(11): 1976.
        4. Stroud N, Mazwi TML, Case LD, et al. Prestroke physical activity and early functional status after stroke. Journal of Neurology, Neurosurgery & Psychiatry 2009;80(9): 1019.
        5. Lee S, Colditz GA, Berkman LF, Kawachi I. Caregiving and risk of coronary heart disease in US women* 1:: A prospective study. American Journal of Preventive Medicine 2003;24(2): 113-9.
        6. Hauser R, Willis R. Survey design and methodology in the Health and Retirement Study and the Wisconsin Longitudinal Study. Population and Development Review. 2004;30:209-235.  (this is a long article – skim it for key features of study design)
        7. Banks J, et al. Disease and Disadvantage in the United States and England. JAMA 2006; 295 (2037)

      •  

      • Pagidipati-749-56 File
        Not available unless: Your ID number contains 02
      • ARIC Designv2 File
        Not available unless: Your ID number contains 02
      • Pickett Inequalities File
        Not available unless: Your ID number contains 02
      • Hauser DesignofHRSandWLS File
        Not available unless: Your ID number contains 02
      • Stroud PreStrokePhysAx File
        Not available unless: Your ID number contains 02
      • Banks DiseaseUSEngland2010 File
        Not available unless: Your ID number contains 02
      • Lee CaregivingCHD File
        Not available unless: Your ID number contains 02

      • Assignment: 

        There are a lot of readings here, but most are quite short. Throughout, the goal is to understand the research design, not the specific content of the study.  Prior to class, please post on the course website answers to the following questions:

        >Choose at least 3 distinct data sources (e.g., ARIC, HRS, death certificate data, NHS, etc), and give an example of a research question (e.g., a hypothesis about the effect of a specific exposure ona  specific outcome) you consider the study exceptionally strong to address.   For each, provide an example of a research question you consider the design very weak to address.  Explain why the data source is strong or weak for each question.  Do not just discuss the questions addressed in the readings, think of new questions, preferably things you might be interested in.  This is not supposed to be a commentary related to the substantive questions in the readings: the goal is to focus on the pros and cons of various data sources. For hypotheses each study would not be well equipped to address, if possible describe another study that could address the hypothesis.

        Extra Assignment for the Applied Data IS:

        Write a summary of the methods and results of your project in the 2000 Census last week. Include a Table 1 describing the data and a Table 2 summarizing the results of your regression models.

    • Monday, April 24, 2017; 10:30 AM - 12:30 PM

      Lecture: Clustered Data: Person level clustering
       This lecture will continue the clustered data discussion from class 2, but instead of focusing on clustering due to spatial autocorrelation, we will discuss clustering due to repeated measures on the same person. This lays the foundation for growth curve analyses of longitudinal change over time. 

      Faculty: Maria Glymour

      Location: Mission Hall 1406

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Wilson RS, Hebert LE, Scherr PA, Barnes LL, Mendes de Leon CF, Evans DA. Educational attainment and cognitive decline in old age. Neurology. 2009;72(5):460.
        2. Hanley et al., Statistical analysis of correlated data using GEE: an orientation.  Am J Epi 2003 v 157, pg 364. 
        3. Hubbard et al To GEE or not to GEE.  Comparing Population Average and Mixed Models for Estimating ASsociations Between Neighborhood Risk Factors and Health. Epidemiology 2010. 
      • wilson edn cogDecline File
        Not available unless: Your ID number contains 02
      • Hubbard GEE File
        Not available unless: Your ID number contains 02
      • Hanley GEE Am. J. Epidemiol.-2003-Hanley-364-75 File
        Not available unless: Your ID number contains 02

      • Optional Reading:  Interaction methods are central to growth curve models.  If you are rusty on interactions, the VanderWeele article is highly recommended.

      • VanderWeele.InteractionTutorial 2014 File
      • Optional reading URL

      • Assignment: 

        Defining the time dimension is a fundamental challenge in longitudinal data analysis.  The most common choice is age or time since study enrollment, however, for many questions other time dimensions are relevant, for example grade in school, or time before or after stroke incidence, or time since release from prison.  Identify an article in the applied literature using a longitudinal analysis based on age as the time dimension, time since study enrollment as the time dimension, and one other possible time dimension (ie, not age or time since study enrollment).  For each study, briefly describe the research question, the study sample, the longitudinal design, and the analysis approach. Please post links to the studies. 

        Additional Assignment for Applied Data Analysis IS: 

        1) Using your Census data set from week 2, estimate the association between age and self-care disability.  Re-estimate the model, controlling for year of birth. Write a paragraph summarizing your analysis approach, findings, and the limitations of this analysis based on cross-sectional data.

        2) Return to the IPUMS site and download the 1990 5% sample, retaining the same core demographics as before and the available disability variables (hint: you can revise your old data request).  

        3) Estimate the same model from step 1 in the 1990 census.  Compare the results.

        4) Pool the two Census data sets. Estimate the same model as in 1 and a new model with an interaction between year of the census and age. Describe your findings.

      • lecture 4 File

    • Monday, May 1, 2017; 10:30 AM - 12:30 PM

      Lecture: Evaluating lifecourse determinants of chronic disease in longitudinal data analysis


      Faculty:  Maria Glymour

      Location:        Mission Hall 1406

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Ben-Shlomo Y, Kuh D. A life course approach to chronic disease epidemiology: conceptual models, empirical challenges and interdisciplinary perspectives. International journal of Epidemiology 2002;31(2):285-293.

        2.   Mishra G, Nitsch D, Black S, De Stavola B, Kuh D, Hardy R. A structured approach to modelling the effects of binary exposure variables over the life course. International journal of epidemiology 2009;38(2):528-537.

        3. Naumova, E., A. Must, et al. (2001). "Tutorial in Biostatistics: Evaluating the impact of critical periods' in longitudinal studies of growth using piecewise mixed effects models." International Journal of Epidemiology 30(6): 1332.

        4. Wills AK, Lawlor DA, Matthews FE, Aihie Sayer A, Bakra E, Ben-Shlomo Y, Benzeval M, Brunner E, Cooper R, Kivimaki M, Kuh D, Muniz-Terrera G, Hardy R. Life Course Trajectories of Systolic Blood Pressure Using Longitudinal Data from Eight UK Cohorts. PLoS Med 2011;8(6):e1000440.

        5. Fitzpatrick A, Kuller L, Lopez O, et al. Midlife and late-life obesity and the risk of dementia: cardiovascular health study. Archives of Neurology. 2009;66(3):336

         

      • Optional Reading:

      • Naumova CriticalPeriods File
        Not available unless: Your ID number contains 02
      • Mishra methods lifecourse File
        Not available unless: Your ID number contains 02
      • Fitzpatrick ObesityDementia File
        Not available unless: Your ID number contains 02
      • Willis LIfecourseBP File
        Not available unless: Your ID number contains 02
      • Ben Shlomo Kuh Lifecourse File
        Not available unless: Your ID number contains 02

      • Assignment: For a specific exposure-outcome combination of interest to you, specify which lifecourse model is likely most appropriate and why you think this is the case. Describe the regression models you could use to test your hypothesis. Are there any possible data sets in which this test could be conducted, and if so, what concerns would you have about interpreting your proposed test of the lifecourse model?

        Optional Additional Assignment for the Applied Data Analysis IS:

        Select a variable that characterizes states that you think potentially related to disability (e.g., state average infant mortality rate; state voter participation rates; state unemployment rate).  Choose a variable for which you can identify a data set with historical trend data for US states. Propose two alternative hypotheses that you can test with your data about potentially relevant exposure ages.  For example: "state infant mortality rate in the year of the individual's birth will predict self-care disability at age 30-40" vs "state infant mortality rate in the year the individual turned 18 will predict self-care disability at age 30-40".  You can alter the age of the outcome or the age of exposure for your hypotheses, but you will probably be constrained by the data sources you are able to access.  Link the external data to your census data and test your two alternative hypotheses.  Describe your hypothesis, analysis approach, and findings. 

      • updated slides class 4 and 5 File

    • Monday, May 8, 2017; 10:30 AM - 12:30 PM

      Mediation and Effect Decomposition
       

      Faculty:  

      Location:        Mission Hall 1406

      • Required Reading:

        1. Judd et al., Dietary Patterns Are Associated With Incident Stroke and Contribute to Excess Risk of Stroke in Black Americans

        2. Nandi, Glymour, Kawachi, and VanderWeele. Using Marginal Structural Models to Estimate the Direct Effect of Adverse Childhood Social Conditions on Onset of Heart Disease, Diabetes, and Stroke. Epidemiology 2012. 

        3. VanderWeele TJ. Mediation Analysis: A Practitioner's Guide. Annual Review of Public Health 2016.

      • Optional Reading:

      • Interaction Tutorial File

      • Assignment:

        Please identify a quantitative research article evaluating mediation in your field and provide the citation.

        What is the primary discipline of the authors?

        Draw a DAG representing the implicit or explicit causal model explored in this paper (you do not need to post your DAG, but we will try to discuss in class).

        What is the exposure of interest?

        What is the outcome of interest?

        What is the hypothesized mediator of interest and how is it measured?

        Describe the modeling approach and briefly report the estimated total, direct, and indirect effects (if these are reported). 

        If the direct effect is reported, would you describe this as a natural direct effect, a controlled direct effect, or something else?

        Do you think there is potential measurement error in the mediator and how would that affect the results?

        Do you think there are unmeasured confounders of the mediator-outcome association and how would that affect the results of the mediation analysis?

        Do you have any critiques of the paper? 

      • Mediation slides File

    • Monday, May 15, 2017; 10:30 AM - 12:30 PM

      Lecture: Evaluating Uncertainty

      We consider confidence intervals, the limits of p-values, the intuition of bootstrapping to estimate confidence intervals, and estimating and interpreting subgroup effects. 

      Faculty:  Maria Glymour


      Location:        Mission Hall 1406  

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Sterne JAC, Smith GD. Sifting the evidence - what's wrong with significance tests? British Medical Journal. Jan 27 2001;322(7280):226-+.
        2.  Ioannidis JPA. Why most published research findings are false. PLoS Med. 2006;2(8):e124: 0696-0701
        3. Grunkemeier, G. and Y. Wu (2004). "Bootstrap resampling methods: something for nothing?" The Annals of thoracic surgery 77(4): 1142.
        4. Ertel et al., Frailty modifies effectiveness of psychosocial intervention in recovery from stroke. Clin Rehabil. 2007. 21:511.

      • Optional Reading:

      • Sterne on p-values File
        Not available unless: Your ID number contains 02
      • On (Ir)Reproducible Science File
        Not available unless: Your ID number contains 02
      • Introduction to the bootstrap File
        Not available unless: Your ID number contains 02
      • Ertel FIRST Secondary File
        Not available unless: Your ID number contains 02

      • Assignment: 

        Specify a hypothesis regarding a particular exposure and outcome and a binary effect modifier including specific measures of association (specify the magnitudes of that association you anticipate: I suggest making everything cross-sectional). Using the software of your choice, generate a population with 1000 people under a causal structure consistent with this hypothesis. Draw a simple random sample 100 individuals from this population and estimate the population average exposure-outcome association and the association stratified by your modifier of interest within this subset.  Repeat this 10 times and write the parameter estimates and CI each time.

        Repeat the data set construction, setting the causal effect to the null.  Again repeat this 10 times and write the parameter estimate and CI each time (if you figure out how to automate it, run it 1000 times and post the histogram of the parameter estimates and p-values).

        Use your code above and also a canned software command to estimate statistical power to detect the difference in means under the settings below:

        *n=100, μ0=.02, μ1=.12, SD=1, α=.05

        *n=100,μ0=.02, μ1=.12, SD=2, α=.05

        *n=500, μ0=.3, μ1=.3, SD=1, α=.05

         For each of the 3 settings above, what is the power to detect whether the ratio of the means=1?

      • Uncertainty Slides File

    • Monday, May 22, 2017; 10:30 AM - 12:30 PM

      Lecture:  Selection Bias and Bias Analyses
       

      Faculty:  Maria Glymour

      Location:  Mission Hall 1406

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        1. Hernán, M.A. & Robins, J.M. (2016). Chapter 8: Selection bias. Causal Inference. Boca Raton: Chapman & Hall/CRC, forthcoming.
        2. Howe, L.D., Tilling, K., Galobardes, B., & Lawlor, D.A. (2013). Loss to follow-up in cohort studies: bias in estimates of socioeconomic inequalities.Epidemiology, 24(1), 1-9.
        3. Lajous, M., Banack, H.R., Kaufman, J.S., & Hernán, M.A. (2015). Should patients with chronic disease be told to gain weight? The obesity paradox and selection bias. The American journal of medicine, 128(4), 334-336.
        4. Whitley Introduction to power and sample size
        5. Normand Tutorial in Biostatistics: Meta-analysis

        6. Ahmed Publication bias in meta-analyses

         

      • Required Readings on Selection Bias Folder
      • Normand intro to meta analyses File
      • Ahmed Publication Bias File
      • Tilling CaptureRecapture File
        Not available unless: Your ID number contains 02
      • Whitley Intro to Sample Size and Power File

      • Optional Reading:

        1. Hernán M.A., Hernández-Díaz S, Robins J.M. A structural approach to selection bias. Epidemiology. 2004;15(5):615-625.
        2. Mayeda E.R., Tchetgen Tchetgen E.J., Power M.C., et al. A simulation platform to quantify survival bias: an application to research on determinants of cognitive decline. American Journal of Epidemiology. In press.
        3. Appendix for Howe et al. paper
      • Optional Readings Folder

      • Assignment: Optional narrative description

      • slides re selection bias File
      • IV slides revise File
      • Selection Bias 2017 File

    • Monday, June 5, 2017; 10:30 AM - 12:30 PM

      Highlighted

      Lecture: Sampling

      We will have a visiting lecturer, Dr. Yea-Hung Chen, who will review approaches to sampling, focusing on approaches to hard-to-reach populations.

      Faculty:  Yea-Hung Chen, PhD

      Location:        Mission Hall 1106

      • Session Slides:

      • Session Audio/Video Recording (Access restricted to registered students):

      • Watch URL

      • Required Reading:

        Required reading:
        • Kandola D, Banner D, O’Keefe-McCarthy S, Jassal D. Sampling methods in cardiovascular nursing research: an overview. Can J Cardiovasc Nurs. 2014; 24(3):15–18.
        • Korn EL, Graubard BI. Epidemiologic studies utilizing surveys: accounting for the sampling design. Am J Public Health. 1991;81(9):1166–1173.
        • Hu SS, Balluz L, Battaglia MP, Frankel MR. Improving public health surveillance using a dual-frame survey of landline and cell phone numbers. Am J Epidemiol. 2011;173(6):703–711.
        • Firestone M, Smylie J, Maracle S, Spiller M, O’Campo P. Unmasking health determinants and health outcomes for urban First Nations using respondent-driven sampling. BMJ Open. 2014;4(7):e004978.
        Optional reading:
        • Langkjær-Bain R. The murky tale of Flint’s deceptive water data. Significance. 2017;14(2):16–21.
        • Lumley T. Analysis of complex survey samples. J Stat Softw. 2004;9(1):1–19.
        • MacKellar DA, Gallagher KM, Finlayson T, Sanchez T, Lansky A, Sullivan PS. Surveillance of HIV risk and prevention behaviors of men who have sex with men—a national application of venue-based, time-space sampling. Public Health Rep. 2007;122(Suppl 1):39–47.
        • Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Philadelphia: Lippincott Williams & Wilkins; 2008: 146–147. [the section on generalizability]
        • Selyukh A. The daredevils without landlines — and why health experts are tracking them. NPR. May 4, 2017.

      • Assignment: 

        It is not unusual to encounter epidemiological or medical studies that use a non-probabilistic sampling scheme. For example, many randomized controlled trials use convenience sampling. Identify a research question of interest to you, the population of interest, and a study design you might use to examine the question. Explain how you might incorporate a sampling strategy into the study design (you might, for example, use respondent-driven sampling to initiate a cohort study). Briefly discuss possible logistical/practical advantages and disadvantages to this plan. Finally, discuss whether you think incorporating the sampling strategy might help (1) reduce bias in the estimation of univariate quantities (such as disease prevalence) and (2) reduce bias in the estimation of causal effects.

      • Kandola File
      • Firestone RDS File
      • Korn Surveys and accounting for sample design File
      • Hu Dual frame survey sampling File
      • Optional Reading: MacKellar Surveillance using time-density space sampling File
      • nhanes stata file
      • Session Slides File
      • NHanes data in csv format File