Identifying mediating variables with graphical models: an application to the study of causal pathways in people living with HIV

Adrian Dobra; Joachim G. Voss; Katherine Buhikire

arxiv: 1907.04838 · v1 · pith:7VRIVQLWnew · submitted 2019-07-10 · 📊 stat.AP · stat.ME

Identifying mediating variables with graphical models: an application to the study of causal pathways in people living with HIV

Adrian Dobra , Katherine Buhikire , Joachim G. Voss This is my paper

Pith reviewed 2026-05-24 23:16 UTC · model grok-4.3

classification 📊 stat.AP stat.ME

keywords graphical modelsmediation analysiscontingency tablesHIVfatigueweaknessloglinear modelscausal pathways

0 comments

The pith

Graphical models on categorical HIV trial data distinguish whether weakness or fatigue mediates treatment effects.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that graphical models recovered from contingency tables can resolve ambiguities in causal mediation that standard methods leave open. In data from people living with HIV, the authors compare two possible mediation directions between treatment, fatigue, and weakness. Standard causal mediation analysis gives inconclusive results, but the graphical approach distinguishes the pathways by examining conditional independencies. This matters because identifying the correct mediator helps target interventions more precisely in clinical settings for symptom management. The work applies the method to a behavioral clinical trial dataset to demonstrate its value.

Core claim

Graphical models fitted to contingency tables via loglinear models can identify mediating variables in causal pathways by encoding conditional independence relationships, allowing distinction between competing mediation structures such as whether weakness mediates treatment effects on fatigue or fatigue mediates effects on weakness, where causal mediation analysis remains inconclusive.

What carries the argument

Graphical models on contingency tables that represent conditional independencies to infer the direction of mediation in categorical data.

If this is right

If the model identifies weakness as the mediator, interventions could target weakness to reduce fatigue levels in PLHIV.
The method supplies a practical alternative for mediation questions in categorical clinical data when traditional analysis fails to decide.
Clarifying mediation directions supports more precise symptom management strategies in HIV behavioral trials.
The approach can be reapplied to similar contingency table datasets from other health studies with binary or categorical outcomes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Incorporating time-ordering from longitudinal measurements could extend the graphs to test mediation over multiple visits.
Sensitivity checks for unmeasured confounding could be added to strengthen confidence in the recovered direction.
The same contingency table approach might map symptom networks in other chronic conditions with overlapping categorical measures.
Validation would require an experiment that directly manipulates the candidate mediator and checks the predicted outcome pattern.

Load-bearing premise

The graphical model recovered from the contingency table correctly encodes the causal direction of mediation without unmeasured confounding or misspecification of the loglinear structure.

What would settle it

A new dataset or reanalysis showing a different pattern of conditional independencies that reverses the identified mediation direction would falsify the specific pathway recovered.

Figures

Figures reproduced from arXiv: 1907.04838 by Adrian Dobra, Joachim G. Voss, Katherine Buhikire.

**Figure 1.** Figure 1: Interaction graph associated with Model 9. This is the graphical loglinear model with minimal sufficient statistics [SSC-W,SSCF][SSC-W,TIME][SSC-W,IC]. Each vertex of this graph corresponds with an observed variable. Each edge of this graph corresponds with a pairwise interaction term of Model 9. interaction term of Model 9, namely, the three pairwise interaction terms [SSC-W,SSCF], [SSC-W,TIME] and [SSC… view at source ↗

**Figure 2.** Figure 2: Out of sample validation for Model 5. Each bar represents the proportion of times based on 10000 sampled datasets Model 5 has been identified to be the loglinear model with the smallest AIC involving variables SSC-W, SSC-F and TIME. From left to right, the bars correspond with sampling proportions of 1%, 2%, . . . , 99%. The horizontal line shows the empirical probability 1/9 = 0.111 of selecting Model 5 … view at source ↗

**Figure 3.** Figure 3: Out of sample validation for Model 9. Each bar represents the proportion of times based on 10000 sampled datasets Model 9 has been identified to be the loglinear model with the smallest AIC involving variables SSC-W, SSC-F, TIME and IC. From left to right, the bars correspond with sampling proportions of 1%, 2%, . . . , 99%. The horizontal line shows the empirical probability 1/114 = 0.0087 of selecting M… view at source ↗

read the original abstract

We empirically demonstrate that graphical models can be a valuable tool in the identification of mediating variables in causal pathways. We make use of graphical models to elucidate the causal pathway through which the treatment influences the levels of fatigue and weakness in people living with HIV (PLHIV) based on a secondary analysis of a categorical dataset collected in a behavioral clinical trial: is weakness a mediator for the treatment and fatigue, or is fatigue a mediator for the treatment and weakness? Causal mediation analysis could not offer any definite answers to these questions.\\ KEYWORDS: Contingency tables; graphical models; loglinear models; HIV; mediation

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Applies standard loglinear graphical models to a 3-way HIV symptom table to pick between two mediation directions, but the undirected structure does not support causal ordering claims without extra assumptions that are not checked.

read the letter

The paper takes an existing loglinear graphical modeling approach for categorical data and applies it to a contingency table from an HIV behavioral trial. The goal is to resolve which symptom mediates the other in the pathway from treatment to fatigue and weakness, after standard causal mediation analysis gave no answer. That is the core of what it does: fit the models to the observed counts and read the conditional independence pattern as evidence for one mediation structure over the other. The example is concrete and the dataset is real, so readers working with similar symptom data might find the illustration useful as a worked case. No new method or framework is added; the contribution is the application itself. The main soft spot is the jump from undirected conditional independence to a specific directed mediation claim. The fitted graph can show, say, that treatment and fatigue are independent given weakness, but that pattern alone does not establish that weakness is the mediator in a causal sense. Doing so requires no unmeasured confounding between any pair and that the selected loglinear terms match the true causal factorization; neither is tested or justified beyond the single observational table. The stress-test note is on target here. The paper is for applied statisticians or HIV researchers who already use graphical models on categorical data and want to see them tried on a mediation question. It is narrow in scope and does not change general practice, but the data example is honest enough that a serious editor should send it to referees rather than desk-reject. The authors engage the literature on why mediation analysis failed and attempt an alternative; that is enough to warrant review even if the causal interpretation needs more support.

Referee Report

1 major / 1 minor

Summary. The manuscript claims to empirically demonstrate that graphical models (via loglinear models on contingency tables) can identify mediating variables in causal pathways. Applied to categorical data from a behavioral clinical trial in people living with HIV, the approach is used to resolve whether weakness mediates treatment effects on fatigue or fatigue mediates treatment effects on weakness, a distinction that standard causal mediation analysis could not make.

Significance. If the method reliably distinguishes mediation directions from observational contingency tables, it could provide a useful complement to directed causal mediation techniques in categorical-data settings. The demonstration rests on recovering conditional independence structure from a single 3-way table and interpreting it as a specific causal ordering.

major comments (1)

[Abstract] Abstract: the claim that the fitted graphical model distinguishes the mediation direction (weakness mediating treatment→fatigue versus fatigue mediating treatment→weakness) is unsupported. Loglinear graphical models are undirected and encode only conditional independencies; mapping this structure to a directed mediation path requires untested assumptions of no unmeasured confounding between all pairs and that the selected loglinear terms match the causal factorization. Neither assumption is stated or validated.

minor comments (1)

The manuscript supplies no equations, model-selection criteria, fitting algorithm, or cross-validation steps for the loglinear graphical model, preventing assessment of reproducibility or sensitivity to specification choices.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and constructive report. We address the single major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the claim that the fitted graphical model distinguishes the mediation direction (weakness mediating treatment→fatigue versus fatigue mediating treatment→weakness) is unsupported. Loglinear graphical models are undirected and encode only conditional independencies; mapping this structure to a directed mediation path requires untested assumptions of no unmeasured confounding between all pairs and that the selected loglinear terms match the causal factorization. Neither assumption is stated or validated.

Authors: We agree that the abstract phrasing is too strong. The fitted loglinear model recovers a conditional independence (treatment ⊥ one symptom | the other) from the observed 3-way table; under the randomized treatment and the assumption that the selected interaction terms correspond to the causal factorization, this structure is consistent with one mediation ordering rather than the other. However, the manuscript does not explicitly list the required no-unmeasured-confounding assumptions between the two symptoms. We will revise the abstract to state that the graphical model identifies the conditional independence supporting a particular ordering (rather than claiming it “distinguishes the mediation direction”), and we will add a dedicated paragraph in the Methods and Discussion sections that states the additional assumptions and notes they are untestable with the available data. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected; standard application of loglinear graphical models to data.

full rationale

The paper applies established loglinear graphical models to a 3-way contingency table from observational data to recover conditional independence structure and interpret it in terms of mediation. No equations, fitted parameters renamed as predictions, or self-citation chains are shown that would reduce the reported mediation direction to a tautology or input by construction. The result is an empirical fit to the observed table under standard model assumptions; the derivation chain does not loop back on itself.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities can be extracted.

pith-pipeline@v0.9.0 · 5633 in / 895 out tokens · 18855 ms · 2026-05-24T23:16:54.072056+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages

[1]

Agresti, Categorical Data Analysis, John Wiley & Sons, New York, 1990

A. Agresti, Categorical Data Analysis, John Wiley & Sons, New York, 1990

work page 1990
[2]

Al-Dakkak, S

I. Al-Dakkak, S. Patel, E. McCann, A. Gadkari, G. Prajapati, and E.M. Maiese, The impact of speciﬁc HIV treatment-related adverse events on adherence to antiretroviral therapy: a systematic review and meta-analysis , AIDS Care 25 (2012), pp. 400–414

work page 2012
[3]

Baron and D.A

R.M. Baron and D.A. Kenny, The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations , Journal of Personality and Social Psychology 51 (1986), pp. 1173–1182

work page 1986
[4]

Barroso, B.G

J. Barroso, B.G. Hammill, J. Leserman, N. Salahuddin, J.L. Harmon, and B.W. Pence, Physiological and psychosocial factors that predict HIV-related fatigue , AIDS and behavior 14 (2010), pp. 1415– 1427

work page 2010
[5]

Barroso and J

J. Barroso and J. Voss, Fatigue in HIV and AIDS: An analysis of evidence, Journal of the Association of Nurses in AIDS Care 24 (2013), pp. S5–14

work page 2013
[6]

Benamar, S

K. Benamar, S. Addou, M. Yondorf, E.B. Geller, T.K. Eisenstein, and M.W. Adler, Intrahypotha- lamic injection of the HIV-1 envelope glycoprotein induces fever via interaction with the chemokine system, The Journal of Pharmacology and Experimental Therapeutics 332 (2010), pp. 549–553

work page 2010
[7]

Bishop, S.E

Y.M.M. Bishop, S.E. Fienberg, and P.W. Holland, Discrete Multivariate Analysis: Theory and Practice, M.I.T. Press, Cambridge, MA, 1975

work page 1975
[8]

Buseh, S.T

A. Buseh, S.T. Kelber, P.E. Stevens, and C.G. Park, Relationship of symptoms, perceived health, and stigma with quality of life among urban HV-infected African American men , Public Health Nursing 25 (2008), pp. 409–419

work page 2008
[9]

Chen, S.Y

W.T. Chen, S.Y. Lee, C.S. Shiu, J.M. Simoni, C. Pan, M. Bao, and H. Lu, Fatigue and sleep dis- turbance in HIV-positive women: a qualitative and biomedical approach , Journal of Clinical Nursing 22 (2012), pp. 1262–1269

work page 2012
[10]

Cook, K.H

P.F. Cook, K.H. Sousa, E.E. Matthews, P.M. Meek, and J. Kwong, Patterns of change in symptom clusters with HIV disease progression , Journal of Pain and Symptom Management 42 (2011), pp. 12–23. IDENTIFYING MEDIATING V ARIABLES WITH GRAPHICAL MODELS 15

work page 2011
[11]

daCosta DiBonaventura, S

M. daCosta DiBonaventura, S. Gupta, M. Cho, and J. Mrus, The association of HIV AIDS treatment side eﬀects with health status, work productivity, and resource use , AIDS Care 24 (2012), pp. 744– 755

work page 2012
[12]

Dalakas, Toxic and drug-induced myopathies, Journal of Neurology, Neurosurgery & Psychiatry 80 (2009), pp

M.C. Dalakas, Toxic and drug-induced myopathies, Journal of Neurology, Neurosurgery & Psychiatry 80 (2009), pp. 832–838

work page 2009
[13]

Dellaportas and J.J

P. Dellaportas and J.J. Forster, Markov chain Monte Carlo model determination for hierarchical and graphical log-linear models, Biometrika 86 (1999), pp. 615–633

work page 1999
[14]

Dellaportas and C

P. Dellaportas and C. Tarantola, Model determination for categorical data with factor level merging , Journal of the Royal Statistical Society: Series B (Statistical Methodology) 67 (2005), pp. 269–283

work page 2005
[15]

Dobra and A

A. Dobra and A. Lenkoski, Copula Gaussian graphical models and their application to modeling functional disability data , Annals of Applied Statistics 5 (2011), pp. 969–993

work page 2011
[16]

Dobra and H

A. Dobra and H. Massam, The mode oriented stochastic search (MOSS) algorithm for log-linear models with conjugate priors , Statistical Methodology 7 (2010), pp. 240–253

work page 2010
[17]

Dobra and A

A. Dobra and A. Mohammadi, Loglinear model selection and human mobility , Annals of Applied Statistics 12 (2018), pp. 815–845

work page 2018
[18]

Duracinsky, S

M. Duracinsky, S. Herrmann, B. Berzins, A. Armstrong, R. Kohli, S.L. Coeur, A. Diouf, I. Fournier, M. Schechter, and O. Chassany,The development of PROQOL-HIV: An international instrument to assess the health-related quality of life of persons living with HIV AIDS, Journal of Acquired Immune Deﬁciency Syndromes 59 (2012), pp. 498–505

work page 2012
[19]

Edwards and T

D.E. Edwards and T. Havranek, A fast procedure for model search in multidimensional contingency tables, Biometrika 72 (1985), pp. 339–351

work page 1985
[20]

Estanislao, D

L. Estanislao, D. Thomas, and D. Simpson, HIV neuromuscular disease and mitochondrial function, Mitochondrion 4 (2004), pp. 131–139

work page 2004
[21]

Fienberg, The analysis of multidimensional contingency tables , Ecology 51 (1970), pp

S.E. Fienberg, The analysis of multidimensional contingency tables , Ecology 51 (1970), pp. 419–433

work page 1970
[22]

Finsterer and S.Z

J. Finsterer and S.Z. Mahjoub, Fatigue in healthy and diseased individuals , American Journal of Hospice and Palliative Medicine 31 (2014), pp. 562–575

work page 2014
[23]

Fritz and D.P

M.S. Fritz and D.P. MacKinnon,A graphical representation of the mediated eﬀect, Behavior Research Methods 40 (2008”), pp. 55–60

work page 2008
[24]

Hidalgo, A

M. Hidalgo, A. Camozzato, L. Cardoso, C. Preussler, C. Nunes, R. Tavares, M. Posser, and M. Chaves, Evaluation of behavioral states among morning and evening active healthy individuals , Brazilian Journal of Medical and Biological Research 35 (2002), pp. 837–842

work page 2002
[25]

Hojsgaard, D

S. Hojsgaard, D. Edwards, and S. Lauritzen, Graphical Models with R , Springer-Verlag, New York, 2012

work page 2012
[26]

K. Imai, L. Keele, and D. Tingley, A general approach to causal mediation analysis , Psychological Methods 15 (2010), pp. 309–334

work page 2010
[27]

K. Imai, D. Tingley, and T. Yamamoto, Experimental designs for identifying causal mechanisms , Journal of the Royal Statistical Society, Series A 176 (2013), pp. 5–51

work page 2013
[28]

Jason and C

L.A. Jason and C. M., Dimensions and assessment of fatigue , in Fatigue Science for Human Health, Y. Yatanabe, B. Evengard, B.H. Natelson, L.A. Jason, and H. Kuratsune, eds., Springer, Tokyo, 2008, pp. 1–16

work page 2008
[29]

Jason, M

L.A. Jason, M. Evans, M. Brown, and N. Porter, What is fatigue? Pathological and nonpathological fatigue, PM&R 2 (2010), pp. 327–331

work page 2010
[30]

Jenkin, T

P. Jenkin, T. Koch, and D. Kralik, The experience of fatigue for adults living with HIV , Journal of Clinical Nursing 15 (2006), pp. 1123–1131

work page 2006
[31]

E. Jong, L. Oudhoﬀ, C. Epskamp, M. Wagener, M. van Duijn, S. Fischer, and E. van Gorp, Pre- dictors and treatment strategies of HIV-related fatigue in the combined antiretroviral therapy era , AIDS 24 (2010), pp. 1387–1405

work page 2010
[32]

Jordan, Graphical models, Statistical Science 19 (2004), pp

M.I. Jordan, Graphical models, Statistical Science 19 (2004), pp. 140–155

work page 2004
[33]

Keyser, Peripheral fatigue: High-energy phosphates and hydrogen ions , PM&R 2 (2010), pp

R.E. Keyser, Peripheral fatigue: High-energy phosphates and hydrogen ions , PM&R 2 (2010), pp. 347–358

work page 2010
[34]

Koller and N

D. Koller and N. Friedman, Probabilistic Graphical Models: Principles and Techniques , Adaptive Computation and Machine Learning series, The MIT Press, 2009

work page 2009
[35]

Lauritzen, Graphical Models, Clarendon Press, Oxford, UK, 1996

S. Lauritzen, Graphical Models, Clarendon Press, Oxford, UK, 1996. 16 ADRIAN DOBRA AND KATHERINE BUHIKIRE AND JOACHIM G. VOSS

work page 1996
[36]

Leavitt and J

V.M. Leavitt and J. DeLuca, Central fatigue: Issues related to cognition, mood and behavior, and psychiatric diagnoses, PM&R 2 (2010), pp. 332–337

work page 2010
[37]

Leserman, J

J. Leserman, J. Barroso, B.W. Pence, N. Salahuddin, and J.L. Harmon, Trauma, stressful life events and depression predict HIV-related fatigue , AIDS care 20 (2008), pp. 1258–1265

work page 2008
[38]

MacKinnon, A.J

D.P. MacKinnon, A.J. Fairchild, and M.S. Fritz, Mediation analysis, Annual Review of Psychology 58 (2007), pp. 593–614

work page 2007
[39]

Madigan and A

D. Madigan and A. Raftery, Model selection and accounting for model uncertainty in graphical models using Occam’s window, Journal of the American Statistical Association 89 (1994), pp. 1535–1546

work page 1994
[40]

Madigan and J

D. Madigan and J. York, Bayesian graphical models for discrete data , International Statistical Review 63 (1995), pp. 215–232

work page 1995
[41]

Madigan and J

D. Madigan and J. York, Bayesian methods for estimation of the size of a closed population , Biometrika 84 (1997), pp. 19–31

work page 1997
[42]

Marconi, B

V.C. Marconi, B. Wu, J. Hampton, C.E. Ord´ o˜ nez, B.A. Johnson, D. Singh, S. John, M. Gordon, A. Hare, R. Murphy, J. Nachega, D.R. Kuritzkes, C. del Rio, H. Sunpath, and South Africa Resis- tance Cohort Study Team Group Authors, Early warning indicators for ﬁrst-line virologic failure independent of adherence measures in a South African urban clinic , AI...

work page 2013
[43]

Nixon, K

S. Nixon, K. O’Brien, R. Glazier, and A.M. Tynan, Aerobic exercise interventions for adults living with HIV/AIDS, Cochrane Database of Systematic Reviews (2005). John Wiley & Sons, Ltd

work page 2005
[44]

¨Onen, A

N.F. ¨Onen, A. Agbebi, E. Shacham, K.E. Stamm, A.R. ¨Onen, and E.T. Overton, Frailty among HIV-infected persons in an urban outpatient care setting, Journal of Infection 59 (2009), pp. 346–352

work page 2009
[45]

Pearl, Causality: Models, Reasoning, and Inference , Cambridge University Press, New York, 2000

J. Pearl, Causality: Models, Reasoning, and Inference , Cambridge University Press, New York, 2000

work page 2000
[46]

Pence, J

B.W. Pence, J. Barroso, J.L. Harmon, J. Leserman, N. Salahuddin, and B.G. Hammill, Chronicity and remission of fatigue in patients with established HIV infection , AIDS patient care and STDs 23 (2009), pp. 239–244

work page 2009
[47]

Powers, Exercise Physiology Theory and Application to Performance , 6th ed., McGraw Hill, Boston, MA, 2007

S.K. Powers, Exercise Physiology Theory and Application to Performance , 6th ed., McGraw Hill, Boston, MA, 2007

work page 2007
[48]

Available at https://www.R-project.org/

R Core Team, R: A Language and Environment for Statistical Computing , R Foundation for Sta- tistical Computing, Vienna, Austria (2018). Available at https://www.R-project.org/

work page 2018
[49]

Salahuddin, J

N. Salahuddin, J. Barroso, J. Leserman, J.L. Harmon, and B.W. Pence, Daytime sleepiness, night- time sleep quality, stressful life events, and HIV-related fatigue , The Journal of the Association of Nurses in AIDS Care 20 (2009), pp. 6–13

work page 2009
[50]

Sliep, M

Y. Sliep, M. Poggenpoel, and A. Gmeiner, The experience of HIV reactive patients in rural Malawi – Part I , Curationis 24 (2001), pp. 56–65

work page 2001
[51]

Tarantola, MCMC model determination for discrete graphical models , Statistical Modelling 4 (2004), pp

C. Tarantola, MCMC model determination for discrete graphical models , Statistical Modelling 4 (2004), pp. 39–61

work page 2004
[52]

Terzian, S

A.S. Terzian, S. Holman, N. Nathwani, E. Robison, K. Weber, M. Young, R.M. Greenblatt, S.J. Gange, and Women’s Interagency HIV Study,Factors associated with preclinical disability and frailty among HIV-infected and HIV-uninfected women in the era of cART , Journal of Women’s Health 18 (2009), pp. 1965–1974

work page 2009
[53]

Tingley, T

D. Tingley, T. Yamamoto, K. Hirose, L. Keele, and K. Imai, mediation: R package for causal mediation analysis, Journal of Statistical Software 59 (2014), pp. 1–38

work page 2014
[54]

Voss, C.J

J.G. Voss, C.J. Portillo, W.L. Holzemer, and M.J. Dodd, Symptom cluster of fatigue and depression in HIV AIDS, Journal of Prevention & Intervention in the Community 33 (2007), pp. 19–34

work page 2007
[55]

Wantland, J.P

D.J. Wantland, J.P. Mullan, W.L. Holzemer, C.J. Portillo, S. Bakken, and E.M. McGhee, Additive eﬀects of numbness and muscle aches on fatigue occurrence in individuals with HIV/AID who are taking antiretroviral therapy, Journal of Pain and Symptom Management 41 (2011), pp. 469–477

work page 2011
[56]

Wantland, W.L

D.J. Wantland, W.L. Holzemer, S. Moezzi, S.S. Willard, J. Arudo, K.M. Kirksey, C.J. Portillo, I.B. Corless, M.E. Rosa, L.L. Robinson, P.K. Nicholas, M.J. Hamilton, E.F. Sefcik, S. Human, M.M. Rivero, M. Maryland, and E. Huang, A randomized controlled trial testing the eﬃcacy of an HIV/AIDS symptom management manual , Journal of Pain and Symptom Management...

work page 2008
[57]

Whittaker, Graphical Models in Applied Multivariate Statistics , John Wiley & Sons, 1990

J. Whittaker, Graphical Models in Applied Multivariate Statistics , John Wiley & Sons, 1990

work page 1990
[58]

asymptomatic

S. Willard, W.L. Holzemer, D.J. Wantland, Y.P. Cuca, K.M. Kirksey, C.J. Portillo, I.B. Corless, M. Rivero-Mndez, M.E. Rosa, P.K. Nicholas, M.J. Hamilton, E. Sefcik, J. Kemppainen, G. Canaval, L. Robinson, S. Moezzi, S. Human, J. Arudo, L.S. Eller, E. Bunch, P.J. Dole, C. Coleman, K. Nokes, N.R. Reynolds, Y.F. Tsai, M. Maryland, J. Voss, and T. Lindgren, D...

work page 2009
[59]

Zhao, J.G

X. Zhao, J.G. Lynch, and Q. Chen, Reconsidering Baron and Kenny: Myths and truths about mediation analysis, Journal of Consumer Research 37 (2010), pp. 197–206. Department of Statistics, University of W ashington, Seattle, W A, USA Frances Payne Bolton School of Nursing, Case Western Reserve University, Cleveland, OH, USA

work page 2010

[1] [1]

Agresti, Categorical Data Analysis, John Wiley & Sons, New York, 1990

A. Agresti, Categorical Data Analysis, John Wiley & Sons, New York, 1990

work page 1990

[2] [2]

Al-Dakkak, S

I. Al-Dakkak, S. Patel, E. McCann, A. Gadkari, G. Prajapati, and E.M. Maiese, The impact of speciﬁc HIV treatment-related adverse events on adherence to antiretroviral therapy: a systematic review and meta-analysis , AIDS Care 25 (2012), pp. 400–414

work page 2012

[3] [3]

Baron and D.A

R.M. Baron and D.A. Kenny, The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations , Journal of Personality and Social Psychology 51 (1986), pp. 1173–1182

work page 1986

[4] [4]

Barroso, B.G

J. Barroso, B.G. Hammill, J. Leserman, N. Salahuddin, J.L. Harmon, and B.W. Pence, Physiological and psychosocial factors that predict HIV-related fatigue , AIDS and behavior 14 (2010), pp. 1415– 1427

work page 2010

[5] [5]

Barroso and J

J. Barroso and J. Voss, Fatigue in HIV and AIDS: An analysis of evidence, Journal of the Association of Nurses in AIDS Care 24 (2013), pp. S5–14

work page 2013

[6] [6]

Benamar, S

K. Benamar, S. Addou, M. Yondorf, E.B. Geller, T.K. Eisenstein, and M.W. Adler, Intrahypotha- lamic injection of the HIV-1 envelope glycoprotein induces fever via interaction with the chemokine system, The Journal of Pharmacology and Experimental Therapeutics 332 (2010), pp. 549–553

work page 2010

[7] [7]

Bishop, S.E

Y.M.M. Bishop, S.E. Fienberg, and P.W. Holland, Discrete Multivariate Analysis: Theory and Practice, M.I.T. Press, Cambridge, MA, 1975

work page 1975

[8] [8]

Buseh, S.T

A. Buseh, S.T. Kelber, P.E. Stevens, and C.G. Park, Relationship of symptoms, perceived health, and stigma with quality of life among urban HV-infected African American men , Public Health Nursing 25 (2008), pp. 409–419

work page 2008

[9] [9]

Chen, S.Y

W.T. Chen, S.Y. Lee, C.S. Shiu, J.M. Simoni, C. Pan, M. Bao, and H. Lu, Fatigue and sleep dis- turbance in HIV-positive women: a qualitative and biomedical approach , Journal of Clinical Nursing 22 (2012), pp. 1262–1269

work page 2012

[10] [10]

Cook, K.H

P.F. Cook, K.H. Sousa, E.E. Matthews, P.M. Meek, and J. Kwong, Patterns of change in symptom clusters with HIV disease progression , Journal of Pain and Symptom Management 42 (2011), pp. 12–23. IDENTIFYING MEDIATING V ARIABLES WITH GRAPHICAL MODELS 15

work page 2011

[11] [11]

daCosta DiBonaventura, S

M. daCosta DiBonaventura, S. Gupta, M. Cho, and J. Mrus, The association of HIV AIDS treatment side eﬀects with health status, work productivity, and resource use , AIDS Care 24 (2012), pp. 744– 755

work page 2012

[12] [12]

Dalakas, Toxic and drug-induced myopathies, Journal of Neurology, Neurosurgery & Psychiatry 80 (2009), pp

M.C. Dalakas, Toxic and drug-induced myopathies, Journal of Neurology, Neurosurgery & Psychiatry 80 (2009), pp. 832–838

work page 2009

[13] [13]

Dellaportas and J.J

P. Dellaportas and J.J. Forster, Markov chain Monte Carlo model determination for hierarchical and graphical log-linear models, Biometrika 86 (1999), pp. 615–633

work page 1999

[14] [14]

Dellaportas and C

P. Dellaportas and C. Tarantola, Model determination for categorical data with factor level merging , Journal of the Royal Statistical Society: Series B (Statistical Methodology) 67 (2005), pp. 269–283

work page 2005

[15] [15]

Dobra and A

A. Dobra and A. Lenkoski, Copula Gaussian graphical models and their application to modeling functional disability data , Annals of Applied Statistics 5 (2011), pp. 969–993

work page 2011

[16] [16]

Dobra and H

A. Dobra and H. Massam, The mode oriented stochastic search (MOSS) algorithm for log-linear models with conjugate priors , Statistical Methodology 7 (2010), pp. 240–253

work page 2010

[17] [17]

Dobra and A

A. Dobra and A. Mohammadi, Loglinear model selection and human mobility , Annals of Applied Statistics 12 (2018), pp. 815–845

work page 2018

[18] [18]

Duracinsky, S

M. Duracinsky, S. Herrmann, B. Berzins, A. Armstrong, R. Kohli, S.L. Coeur, A. Diouf, I. Fournier, M. Schechter, and O. Chassany,The development of PROQOL-HIV: An international instrument to assess the health-related quality of life of persons living with HIV AIDS, Journal of Acquired Immune Deﬁciency Syndromes 59 (2012), pp. 498–505

work page 2012

[19] [19]

Edwards and T

D.E. Edwards and T. Havranek, A fast procedure for model search in multidimensional contingency tables, Biometrika 72 (1985), pp. 339–351

work page 1985

[20] [20]

Estanislao, D

L. Estanislao, D. Thomas, and D. Simpson, HIV neuromuscular disease and mitochondrial function, Mitochondrion 4 (2004), pp. 131–139

work page 2004

[21] [21]

Fienberg, The analysis of multidimensional contingency tables , Ecology 51 (1970), pp

S.E. Fienberg, The analysis of multidimensional contingency tables , Ecology 51 (1970), pp. 419–433

work page 1970

[22] [22]

Finsterer and S.Z

J. Finsterer and S.Z. Mahjoub, Fatigue in healthy and diseased individuals , American Journal of Hospice and Palliative Medicine 31 (2014), pp. 562–575

work page 2014

[23] [23]

Fritz and D.P

M.S. Fritz and D.P. MacKinnon,A graphical representation of the mediated eﬀect, Behavior Research Methods 40 (2008”), pp. 55–60

work page 2008

[24] [24]

Hidalgo, A

M. Hidalgo, A. Camozzato, L. Cardoso, C. Preussler, C. Nunes, R. Tavares, M. Posser, and M. Chaves, Evaluation of behavioral states among morning and evening active healthy individuals , Brazilian Journal of Medical and Biological Research 35 (2002), pp. 837–842

work page 2002

[25] [25]

Hojsgaard, D

S. Hojsgaard, D. Edwards, and S. Lauritzen, Graphical Models with R , Springer-Verlag, New York, 2012

work page 2012

[26] [26]

K. Imai, L. Keele, and D. Tingley, A general approach to causal mediation analysis , Psychological Methods 15 (2010), pp. 309–334

work page 2010

[27] [27]

K. Imai, D. Tingley, and T. Yamamoto, Experimental designs for identifying causal mechanisms , Journal of the Royal Statistical Society, Series A 176 (2013), pp. 5–51

work page 2013

[28] [28]

Jason and C

L.A. Jason and C. M., Dimensions and assessment of fatigue , in Fatigue Science for Human Health, Y. Yatanabe, B. Evengard, B.H. Natelson, L.A. Jason, and H. Kuratsune, eds., Springer, Tokyo, 2008, pp. 1–16

work page 2008

[29] [29]

Jason, M

L.A. Jason, M. Evans, M. Brown, and N. Porter, What is fatigue? Pathological and nonpathological fatigue, PM&R 2 (2010), pp. 327–331

work page 2010

[30] [30]

Jenkin, T

P. Jenkin, T. Koch, and D. Kralik, The experience of fatigue for adults living with HIV , Journal of Clinical Nursing 15 (2006), pp. 1123–1131

work page 2006

[31] [31]

E. Jong, L. Oudhoﬀ, C. Epskamp, M. Wagener, M. van Duijn, S. Fischer, and E. van Gorp, Pre- dictors and treatment strategies of HIV-related fatigue in the combined antiretroviral therapy era , AIDS 24 (2010), pp. 1387–1405

work page 2010

[32] [32]

Jordan, Graphical models, Statistical Science 19 (2004), pp

M.I. Jordan, Graphical models, Statistical Science 19 (2004), pp. 140–155

work page 2004

[33] [33]

Keyser, Peripheral fatigue: High-energy phosphates and hydrogen ions , PM&R 2 (2010), pp

R.E. Keyser, Peripheral fatigue: High-energy phosphates and hydrogen ions , PM&R 2 (2010), pp. 347–358

work page 2010

[34] [34]

Koller and N

D. Koller and N. Friedman, Probabilistic Graphical Models: Principles and Techniques , Adaptive Computation and Machine Learning series, The MIT Press, 2009

work page 2009

[35] [35]

Lauritzen, Graphical Models, Clarendon Press, Oxford, UK, 1996

S. Lauritzen, Graphical Models, Clarendon Press, Oxford, UK, 1996. 16 ADRIAN DOBRA AND KATHERINE BUHIKIRE AND JOACHIM G. VOSS

work page 1996

[36] [36]

Leavitt and J

V.M. Leavitt and J. DeLuca, Central fatigue: Issues related to cognition, mood and behavior, and psychiatric diagnoses, PM&R 2 (2010), pp. 332–337

work page 2010

[37] [37]

Leserman, J

J. Leserman, J. Barroso, B.W. Pence, N. Salahuddin, and J.L. Harmon, Trauma, stressful life events and depression predict HIV-related fatigue , AIDS care 20 (2008), pp. 1258–1265

work page 2008

[38] [38]

MacKinnon, A.J

D.P. MacKinnon, A.J. Fairchild, and M.S. Fritz, Mediation analysis, Annual Review of Psychology 58 (2007), pp. 593–614

work page 2007

[39] [39]

Madigan and A

D. Madigan and A. Raftery, Model selection and accounting for model uncertainty in graphical models using Occam’s window, Journal of the American Statistical Association 89 (1994), pp. 1535–1546

work page 1994

[40] [40]

Madigan and J

D. Madigan and J. York, Bayesian graphical models for discrete data , International Statistical Review 63 (1995), pp. 215–232

work page 1995

[41] [41]

Madigan and J

D. Madigan and J. York, Bayesian methods for estimation of the size of a closed population , Biometrika 84 (1997), pp. 19–31

work page 1997

[42] [42]

Marconi, B

V.C. Marconi, B. Wu, J. Hampton, C.E. Ord´ o˜ nez, B.A. Johnson, D. Singh, S. John, M. Gordon, A. Hare, R. Murphy, J. Nachega, D.R. Kuritzkes, C. del Rio, H. Sunpath, and South Africa Resis- tance Cohort Study Team Group Authors, Early warning indicators for ﬁrst-line virologic failure independent of adherence measures in a South African urban clinic , AI...

work page 2013

[43] [43]

Nixon, K

S. Nixon, K. O’Brien, R. Glazier, and A.M. Tynan, Aerobic exercise interventions for adults living with HIV/AIDS, Cochrane Database of Systematic Reviews (2005). John Wiley & Sons, Ltd

work page 2005

[44] [44]

¨Onen, A

N.F. ¨Onen, A. Agbebi, E. Shacham, K.E. Stamm, A.R. ¨Onen, and E.T. Overton, Frailty among HIV-infected persons in an urban outpatient care setting, Journal of Infection 59 (2009), pp. 346–352

work page 2009

[45] [45]

Pearl, Causality: Models, Reasoning, and Inference , Cambridge University Press, New York, 2000

J. Pearl, Causality: Models, Reasoning, and Inference , Cambridge University Press, New York, 2000

work page 2000

[46] [46]

Pence, J

B.W. Pence, J. Barroso, J.L. Harmon, J. Leserman, N. Salahuddin, and B.G. Hammill, Chronicity and remission of fatigue in patients with established HIV infection , AIDS patient care and STDs 23 (2009), pp. 239–244

work page 2009

[47] [47]

Powers, Exercise Physiology Theory and Application to Performance , 6th ed., McGraw Hill, Boston, MA, 2007

S.K. Powers, Exercise Physiology Theory and Application to Performance , 6th ed., McGraw Hill, Boston, MA, 2007

work page 2007

[48] [48]

Available at https://www.R-project.org/

R Core Team, R: A Language and Environment for Statistical Computing , R Foundation for Sta- tistical Computing, Vienna, Austria (2018). Available at https://www.R-project.org/

work page 2018

[49] [49]

Salahuddin, J

N. Salahuddin, J. Barroso, J. Leserman, J.L. Harmon, and B.W. Pence, Daytime sleepiness, night- time sleep quality, stressful life events, and HIV-related fatigue , The Journal of the Association of Nurses in AIDS Care 20 (2009), pp. 6–13

work page 2009

[50] [50]

Sliep, M

Y. Sliep, M. Poggenpoel, and A. Gmeiner, The experience of HIV reactive patients in rural Malawi – Part I , Curationis 24 (2001), pp. 56–65

work page 2001

[51] [51]

Tarantola, MCMC model determination for discrete graphical models , Statistical Modelling 4 (2004), pp

C. Tarantola, MCMC model determination for discrete graphical models , Statistical Modelling 4 (2004), pp. 39–61

work page 2004

[52] [52]

Terzian, S

A.S. Terzian, S. Holman, N. Nathwani, E. Robison, K. Weber, M. Young, R.M. Greenblatt, S.J. Gange, and Women’s Interagency HIV Study,Factors associated with preclinical disability and frailty among HIV-infected and HIV-uninfected women in the era of cART , Journal of Women’s Health 18 (2009), pp. 1965–1974

work page 2009

[53] [53]

Tingley, T

D. Tingley, T. Yamamoto, K. Hirose, L. Keele, and K. Imai, mediation: R package for causal mediation analysis, Journal of Statistical Software 59 (2014), pp. 1–38

work page 2014

[54] [54]

Voss, C.J

J.G. Voss, C.J. Portillo, W.L. Holzemer, and M.J. Dodd, Symptom cluster of fatigue and depression in HIV AIDS, Journal of Prevention & Intervention in the Community 33 (2007), pp. 19–34

work page 2007

[55] [55]

Wantland, J.P

D.J. Wantland, J.P. Mullan, W.L. Holzemer, C.J. Portillo, S. Bakken, and E.M. McGhee, Additive eﬀects of numbness and muscle aches on fatigue occurrence in individuals with HIV/AID who are taking antiretroviral therapy, Journal of Pain and Symptom Management 41 (2011), pp. 469–477

work page 2011

[56] [56]

Wantland, W.L

D.J. Wantland, W.L. Holzemer, S. Moezzi, S.S. Willard, J. Arudo, K.M. Kirksey, C.J. Portillo, I.B. Corless, M.E. Rosa, L.L. Robinson, P.K. Nicholas, M.J. Hamilton, E.F. Sefcik, S. Human, M.M. Rivero, M. Maryland, and E. Huang, A randomized controlled trial testing the eﬃcacy of an HIV/AIDS symptom management manual , Journal of Pain and Symptom Management...

work page 2008

[57] [57]

Whittaker, Graphical Models in Applied Multivariate Statistics , John Wiley & Sons, 1990

J. Whittaker, Graphical Models in Applied Multivariate Statistics , John Wiley & Sons, 1990

work page 1990

[58] [58]

asymptomatic

S. Willard, W.L. Holzemer, D.J. Wantland, Y.P. Cuca, K.M. Kirksey, C.J. Portillo, I.B. Corless, M. Rivero-Mndez, M.E. Rosa, P.K. Nicholas, M.J. Hamilton, E. Sefcik, J. Kemppainen, G. Canaval, L. Robinson, S. Moezzi, S. Human, J. Arudo, L.S. Eller, E. Bunch, P.J. Dole, C. Coleman, K. Nokes, N.R. Reynolds, Y.F. Tsai, M. Maryland, J. Voss, and T. Lindgren, D...

work page 2009

[59] [59]

Zhao, J.G

X. Zhao, J.G. Lynch, and Q. Chen, Reconsidering Baron and Kenny: Myths and truths about mediation analysis, Journal of Consumer Research 37 (2010), pp. 197–206. Department of Statistics, University of W ashington, Seattle, W A, USA Frances Payne Bolton School of Nursing, Case Western Reserve University, Cleveland, OH, USA

work page 2010