1. Part 2 Attrition: Bias and Loss of Power

2. Relevant Papers Graham, J.W., (2009). Missing data analysis: making it work in the real world. Annual Review of Psychology, 60, 549-576. Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychological Methods, 6, 330_351. Hedeker, D., & Gibbons, R.D. (1997). Application of random-effects pattern-mixture models for missing data in longitudinal studies, Psychological Methods, 2, 64-78. Graham, J.W., & Collins, L.M. (2010, forthcoming). Using Modern Missing Data Methods with Auxiliary Variables to Mitigate the Effects of Attrition on Statistical Power. Chapter 10 in Graham (2010, forthcoming), Missing Data: Analysis and Design. New York: Springer.

3. Relevant Papers Graham, J.W., Palen, L.A., et al. (2008). Attrition: MAR & MNAR missingness, and estimation bias. Annual Meetings of the Society for Prevention Research, San Francisco, CA. (available upon request) also see: Graham, J.W., (2010, forthcoming). Simulations with Missing Data. Chapter 9 in Graham (2010, forthcoming), Missing Data: Analysis and Design. New York: Springer.

4. What if the cause of missingness is MNAR? Problems with this statement MAR & MNAR are widely misunderstood concepts I argue that the cause of missingness is never purely MNAR The cause of missingness is virtually never purely MAR either.

5. MAR vs MNAR "Pure" MCAR, MAR, MNAR never occur in field research Each requires untenable assumptions e.g., that all possible correlations and partial correlations are r = 0

6. MAR vs MNAR Better to think of MAR and MNAR as forming a continuum MAR vs MNAR NOT even the dimension of interest

7. MAR vs MNAR: What IS the Dimension of Interest? How much estimation bias? when cause of missingness cannot be included in the model

8. Bottom Line... All missing data situations are partly MAR and partly MNAR Sometimes it matters... bias affects statistical conclusions Often it does not matter bias has tolerably little effect on statistical conclusions (Collins, Schafer, & Kam, Psych Methods, 2001)

9. Methods: "Old" vs MAR vs MNAR MAR methods (MI and ML) are ALWAYS at least as good as, usually better than "old" methods (e.g., listwise deletion) Methods designed to handle MNAR missingness are NOT always better than MAR methods

10. Yardstick for Measuring Bias Standardized Bias = (average parameter est) – (population value) -------------------------------------------------------- X 100 Standard Error (SE) |bias| < 40 considered small enough to be tolerable t-value off by 0.4

11. A little background for Collins, Schafer, & Kam (2001; CSK) Example model of interest: X  Y X = Program (prog vs control) Y = Cigarette Smoking Z = Cause of missingness: say, Rebelliousness (or smoking itself) Factors to be considered: % Missing (e.g., % attrition) r YZ r ZR

12. r YZ Correlation between cause of missingness (Z) e.g., rebelliousness (or smoking itself) and the variable of interest (Y) e.g., Cigarette Smoking

13. rZRrZR Correlation between cause of missingness (Z) e.g., rebelliousness (or smoking itself) and missingness on variable of interest e.g., Missingness on the Smoking variable Missingness on Smoking (often designated: R or R Y ) Dichotomous variable: R = 1: Smoking variable not missing R = 0: Smoking variable missing

14. CSK Study Design (partial) Simulations manipulated amount of missingness (25% vs 50%) r ZY (r =.40, r =.90) r ZR held constant r =.45 with 50% missing (applies to "MNAR-Linear" missingness)

15. CSK Results (partial) (MNAR Missingness) 25% missing, r YZ =.40... no problem 25% missing, r YZ =.90... no problem 50% missing, r YZ =.40... no problem 50% missing, r YZ =.90... problem * "no problem" = bias does not interfere with inference These Results apply to the regression coefficient for X  Y with "MNAR-Linear" missingness (see CSK, 2001, Table 2)

16. But Even CSK Results Too Conservative Not considered by CSK: r ZR In their simulation r ZR =.45 Even with 50% missing and r YZ =.90 bias can be acceptably small Graham et al. (2008): Bias acceptably small (standardized bias < 40) as long as r ZR <.24

17. r ZR <.24 Very Plausible Study r ZR ______________ HealthWise (Caldwell, Smith, et al., 2004).106 AAPT (Hansen & Graham, 1991).093 Botvin1.044 Botvin2.078 Botvin3.104 All of these yield standardized bias < 10 (estimated)

18. Attrition in HealthWise Best (pretest) predictors of Attrition in HW Gender Lifetime Sex Lifetime Alcohol Use Lifetime Smoking Lifetime Dagga Use

19. CSK and Follow-up Simulations Results very promising Suggest that even MNAR biases are often tolerably small But these simulations still too narrow

20. Beginnings of a Taxonomy of Attrition Causes of Attrition on Y (main DV) Case 1: not Program (P), not Y, not PY interaction Case 2: P only Case 3: Y only... (CSK scenario) Case 4: P and Y only Graham, J. W. (2009). Annual Review of Psychology.

21. Beginnings of a Taxonomy of Attrition Causes of Attrition on Y (main DV) Case 5: PY interaction only Case 6: P + PY interaction Case 7: Y + PY interaction Case 8: P, Y, and PY interaction

22. Taxonomy of Attrition Cases 1-4 often little or no problem Cases 5-8 Jury still out (more research needed) Very likely not as much of a problem as previously though Use diagnostics to shed light

23. Use of Missing Data Diagnostics Diagnostics based on pretest data not much help Hard to predict missing distal outcomes from differences on pretest scores Longitudinal Diagnostics can be much more helpful

24. Hedeker & Gibbons (1997) Plot main DV over time for four groups: for Program and Control for those with and without last wave of data Much can be learned

25. Empirical Examples Hedeker & Gibbons (1997) Drug treatment of psychiatric patients Hansen & Graham (1991) Adolescent Alcohol Prevention Trial (AAPT) Alcohol, smoking, other drug prevention among normal adolescents (7 th – 11 th grade)

26. Empirical Example Used by Hedeker & Gibbons (1997) IV: Drug Treatment vs. Placebo Control DV: I npatient M ultidimensional P sychiatric S cale (IMPS) 1 = normal 2 = borderline mentally ill 3 = mildly ill 4 = moderately ill 5 = markedly ill 6 = severely ill 7 = among the most extremely ill

27. From Hedeker & Gibbons (1997) IMPS low = better outcomes Placebo Control Drug Treatment Weeks of Treatment

28. Longitudinal Diagnostics Hedeker & Gibbons Example Treatment droppers do BETTER than stayers Control droppers do WORSE than stayers Example of Program X DV interaction But in this case, pattern would lead to suppression bias Not as bad for internal validity in presence of significant program effect

29. AAPT ( Hansen & Graham, 1991) IV: Normative Education Program vs Information Only Control DV: Cigarette Smoking (3-item scale) Measured at one-year intervals 7 th grade – 11 th grade

30. AAPT Cigarette Smoking (high = more smoking; arbitrary scale) th Control Program

31. Longitudinal Diagnostics AAPT Example Treatment droppers do WORSE than stayers little steeper increase Control droppers do WORSE than stayers little steeper increase Little evidence for Prog X DV interaction Very likely MAR methods allow good conclusions (CSK scenario holds)

32. Use of Auxiliary Variables Reduces attrition bias Restores some power lost due to attrition

33. What Is an Auxiliary Variable? A variable correlated with the variables in your model but not part of the model not necessarily related to missingness used to "help" with missing data estimation Best auxiliary variables: same variable as main DV, but measured at waves not used in analysis model

34. Model of Interest

35. Benefit of Auxiliary Variables Example from Graham & Collins (2010) X Y Z 1 1 1 500 complete cases 1 0 1500 cases missing Y X, Y variables in the model (Y sometimes missing) Z is auxiliary variable

36. Benefit of Auxiliary Variables Effective sample size (N') Analysis involving N cases, with auxiliary variable(s) gives statistical power equivalent to N' complete cases without auxiliary variables

37. Benefit of Auxiliary Variables It matters how highly Y and Z (the auxiliary variable) are correlated For example increase r YZ =.40 N = 500 gives power of N' = 542( 8%) r YZ =.60 N = 500 gives power of N' = 608 (22%) r YZ =.80 N = 500 gives power of N' = 733(47%) r YZ =.90 N = 500 gives power of N' = 839(68%)

38. Effective Sample Size by r YZ r YZ Effective Sample Size

39. Conclusions Attrition CAN be bad for internal validity But often it's NOT nearly as bad as often feared Don't rush to conclusions, even with rather substantial attrition Examine evidence (especially longitudinal diagnostics) before drawing conclusions Use MI and ML missing data procedures! Use good auxiliary variables to minimize impact of attrition

40. Part 3: Illustration of Missing Data Analysis: Multiple Imputation with NORM and Proc MI

41. Multiple Imputation: Basic Steps Impute Analyze Combine results

42. Imputation and Analysis Impute 40 datasets a missing value gets a different imputed value in each dataset Analyze each data set with USUAL procedures e.g., SAS, SPSS, LISREL, EQS, STATA, HLM Save parameter estimates and SE’s

43. Combine the Results Parameter Estimates to Report Average of estimate (b-weight) over 40 imputed datasets

44. Combine the Results Standard Errors to Report Weighted sum of: “within imputation” variance average squared standard error usual kind of variability “between imputation” variance sample variance of parameter estimates over 40 datasets variability due to missing data

45. Materials for SPSS Regression Starting place http://methodology.psu.edu downloads (you will need to get a free user ID to download all our free software) missing data software Joe Schafer's Missing Data Programs John Graham's Additional NORM Utilities http://mcgee.hhdev.psu.edu/missing/index.html (this mcgee website is currently down, but I hope to have it up again in the Fall). Please email me with any questions.

46. exit for sample analysis