DAGitty User Manual
Contents
1. 3 4 Deleting connections A connection is deleted just like it has been inserted i e by double clicking first on the start variable and then on the target variable A connection is also deleted automatically if a new one is inserted in the opposite direction see above 3 5 Setting exposition and outcome As explained above per default the exposition is the variable in the first line of the variable list and the outcome is the one on the second line To turn a different variable into the exposition move the mouse pointer over that variable and hit the e key for the outcome hit the o key instead Doing so will change the colors of the vertices on the canvas to reflect the new structure of the graph 3 6 Workarounds for functions that are still missing Some functions are not yet there in DAGitty but would be nice to have and shall be implemented in future versions In the meantime the following workarounds can be used e Renaming variables This is not yet conveniently possible However you can copy amp paste the vertex labels and adjacency list to a word processor of your choice and then replace every occurence of the variable name of choice with the new version ussing the word processor s search and replace functions Afterwards copy the model description back into DAGitty 4 Adjustment sets Finding sufficient adjustment sets is one main purpose of DAGitty In a nutshell a sufficient adjustment set is a set S of covariates s2. sets and display them in the List of minimal sufficient adjustment sets text field 4 3 Verfiying that all paths are blocked in small models For small models DAGitty will list relevant open and closed paths in the Open and closed paths text field so you can verify that the listed adjustment sets are indeed sufficient if you don t trust DAGitty by checking if every path is indeed blocked For larger models only up to 100 paths each will be listed the list of paths grows exponen tially and becomes too large to fit in computer memory let alone to be verified by hand DAGitty will indicate that it has cancelled the search for more paths by putting at the end of the list Remember also that DAGitty will not list paths that contain a non ancestor of exposure and outcome i e a node colored in gray for the reasons mentioned above 4 4 Adjusting for specific covariates You can also tell DAGitty that you wish a specific covariate to be included into every adjustment set To do this move the mouse over the vertex of that covariate and press the a key DAGitty will then update the list of minimal sufficient adjustment sets accordingly every set displayed is now minimal in the sense that removing any vertex except those you specified will render that set insufficient However DAGitty will refuse to adjust for a variable that is a successor of the exposure see above 5 Acknowledgements The author wishes to than
3. DAGitty User Manual Johannes Textor October 13 2010 Abstract DAGitty is a program for creating editing and analyzing causal models known in epi demiology as directed acyclic graphs DAGs The main task of the program is to assist the user in identifying adjustment sets that is sets of covariates to adjust for in order to isolate the causal effects from an exposure to an outcome from the non causal or confounded effects DAGitty runs in any web browser that supports modern HTML and JavaScript Contents 1 1 Causal models s sa sasra nadaa a re oe 2 Loading and saving models 2 1 DAGitty s textual syntax for causal models 2 2 Loading a model into DAGitty 2 3 Modifying the graphical layout of a model 2 4 Saving the model 3 Editing models within DAGitty u N Set 3 5 Setting exposition and outcome 3 6 Workarounds for functions that are still missing 4 Adjustment sets 4 1 Minimal sufficient adjustment sets 4 2 Finding minimal sufficient adjustment sets 4 3 Verfiying that all paths are blocked in small models 4 4 Adjusting for specific covariates Acknowledgements Legal notice 7 Bundled libraries 8 Bundled examples 9 Author contact oo 1fr fh Awww wnmnn nd DAO oo a 1 Introduction 1 1 Causal models To convey an idea of the purpose of DAGitty this introduction contains some very small examples of causal models confounding and
4. adjustment sets for a more detailed discussion of these subjects we recommend the book Causality by Judea Pear 6 Causal models are also called Bayesian networks in computer science or even DAGs in epidemiology Simply put a DAG is a formal model about causal relationships between certain entities of interest in a specific scenario For example the sentence smoking causes cancer could be translated into the following simple causal model smoking cancer Figure 1 A very simple causal model An important application for causal models which is also the focus of DAGitty is to isolate the causal effects of a variable of interest called exposure onto another called outcome from the confounded relations between the two variables For example consider the following slightly more complex causal model smoking a carry matches gt cancer Figure 2 A classical confounding triangle If we were to perform a study on the relationship between carrying matches in one s pocket and developing lung cancer we would probably find a correlation between these two variables However as the above model indicates this correlation would not imply that carrying matches in your pocket causes lung cancer Smokers are more likely to carry matches in their pockets and also more likely to develop lung cancer This is a classical example of a confounded association between two variables In this example would
5. bine Schipf Robin Haring Nele Friedrich Matthias Nauck Katharina Lau Dietrich Alte Andreas Stang Henry V lzke and Henri Wallaschofski Low total testosterone is associated with increased risk of incident type 2 diabetes mellitus in men Results from the study of health in pomerania SHIP The Aging Male 2010 in press Ian Shrier and Robert W Platt Reducing bias through directed acyclic graphs BMC Medical Research Methodology 8 70 2008 Philipp Strathausen Dracula graph layout and drawing framework 2010 Prototype Core Team Prototype http www prototypejs org 2010 Jin Tian Azaria Paz and Judea Pearl Finding minimal d separators Technical Report R 254 1998
6. ete vertices and connections on the canvas itself All such changes to the model are immediately reflected in the vertex labels and adjacency list text fields Furthermore the list of minimal sufficient adjustment sets see next section will be updated 3 1 Adding new variables To add a new variable to the model double click on a free space in the canvas i e not on an existing variable or press the n key A small dialog will pop up asking you for the name of the new variable Enter the name into the dialog and press the enter key or click OK If you click Cancel no new variable will be created 3 2 Adding new connections To add a new connection double click first on the source vertex which will become highlighted and then on the target vertex The connection will be inserted If a connection existed before in the opposite direction that connection will be deleted because otherwise there would now be a cycle in the model Instead of double clicking on a vertex you can also move the mouse pointer over the vertex and press the key c 3 3 Deleting variables To delete a variable move the mouse pointer over that variable and hit the del key on your keyboard All connections to that variable will be deleted along with the variable DAGitty will refuse to delete the exposition or the outcome variable from the model if you wish to do so you must previously select a new exposition outcome see below
7. his manual 1 3 Installing DAGitty on your own computer DAGitty can be installed on your computer for use without an internet connection To do this download the file which is a ZIP archive containing DAGitty s source Unpack this ZIP file anywhere on your local hard drive To run DAGitty just open the file dags htm1 in the unpacked folder 2 Loading and saving models This section covers the three basic steps of working with DAGitty 1 Loading a model 2 manipulating the graphical layout of the model and 3 saving the model First of all any causal model consists of vertices variables and edges relationships between variables You can either create the model in a text editor such as Microsoft Word Q and then import this description into DAGitty or create the model in DAGitty itself using the graphical user interface see next section In addition DAGitty contains some pre defined examples that you can use to become familiar with the program To do so select one of the pre define examples from the drop down menu below the legend and click on draw DAG 2 1 DAGitty s textual syntax for causal models DAGitty s textual syntax for causal models is compatible with the one used by the DAG program by Sven Kniippel 4 A model description consists of two parts 1 A list of the variables in the model 2 A list of connections between the variables The list of variables is simply one variable per line blank line
8. ins some builtin examples for didactic and illustrative purposes Some of these examples are taken from published papers or talks given at scientific meetings These are in inverse chronological order e Polzer et al 2010 e Schipf et al 2010 e Shrier amp Pratt 2008 e Aicd amp Campos 1996 9 Author contact The author of DAGitty i e me would be glad to receive feedback from those who use DAGitty in their research or for educational purpose Also you can E mail me with suggestions or requests for features that you miss in DAGitty Johannes Textor Institut ftir Theoretische Informatik University of L beck Germany textor tcs uni luebeck de www tcs uni luebeck de mitarbeiter textor References 1 Silvia Acid and Luis M De Campos An algorithm for finding minimum d separating sets in belief networks In Proceedings of the twelfth Conference of Uncertainty in Artificial Intelli gence pages 3 10 1996 Dmitry Baranovskiy Raphael javascript library nttp raphaeljs com 2010 Ines Polzer et al 2010 personal communication Sven Kniippel and Andreas Stang DAG program identifying minimal sufficient adjustment sets Epidemiology Cambridge Mass 21 1 159 2010 S L Laurizen A P Dawid B N Larsen and H G Leimer Independence properties of directed markov fields Networks 20 5 491 505 1990 Judea Pearl Causality models reasoning and inference Cambridge University Press 2000 Sa
9. k Michael Elberfeld Juliane Hardt Sven Knuppel and Sabine Schipf in alphabetical order for enlightening discussions about DAGs that made this program possible 6 Legal notice Use of DAGitty is of course freely permitted and free of charge You may download a copy of DAGitty s source code from its website at www tcs uni luebeck de sonderseiten software y The source code is available under the GNU General Public License GPL either version 2 0 or any later version at the licensee s choice see the file LICENSE txt in the download archive for details In particular the GPL permits you to modify and redistribute the source as you please as long as the result remains itself under the GPL 7 Bundled libraries DAGitty ships along with the following JavaScript libraries e Rapha l a library for smooth cross browser vector graphics in SVG and VML developed by Dmitry Baranovskiy and licensed under the MIT license 2 e Prototype js a framework that makes life with JavaScript much easier Only some parts of Prototype mainly those focusing on data structures are included to keep the code small Developed by the Prototype Core Team and licensed under the MIT license 10 Furthermore DAGitty uses some modified code from the Dracula Graph Library by Philipp Strathausen which is also licensed under the MIT license 9 I am grateful to all authors of these libraries for their valuable work 8 Bundled examples DAGitty conta
10. nts it with additional position information vertex labels adjacency list resulting graph augmented vertex labels E ED A B E 1 2 2 1 6 D AEZ D 1 1 4 1 6 A BDZ A 7 a A 1 2 2 1 5 B ZED B 1 1 4 1 5 Z A N Z 1 0 3 0 1 E D Figure 3 Example for a textual model definition with DAGitty When the model is edited within DAGitty the vertex labels are augmented with additional information that DAGitty uses to layout the vertices on the canvas rightmost column In the second column weights are given for each variable not used yet but perhaps in future versions of DAGitty and in the third column the layout coordinates of each variable are indicated behind the sign for each vertex In general all changes you make to your model within DAGitty are immediately reflected in the vertex labels and adjacency list text fields 2 4 Saving the model To save your model locally just copy amp paste the contents of the vertex labels and adjacency list text fields to a text file e g a Microsoft Word document and save that file locally to your computer Next time you wish to work on the model copy the model description back into DAGitty as explained above 3 Editing models within DAGitty As explained above you are free to make changes directly to the textual description of your model which will be reflected on the canvas next time you click on draw DAG However you can also create modify and del
11. re outcome or both see the legend on the left hand side of the screen To give you an idea of the model s complexity DAGitty will count all open paths but not the closed ones and display this information below the legend 4 1 Minimal sufficient adjustment sets A minimal sufficient adjustment set MSAS is a sufficient adjustment set of which no proper subset is itself sufficient For example consider again the causal model in Figure In this example the following three sets are sufficient adjustment sets A B Z A Z B Z Of these three sets A Z and B Z are minimal sufficient adjustment sets while the set A B Z is sufficient but not minimal Note that adjusting for Z is not sufficient since this would open the path E A gt Z B D Since both E and D depend on Z adjusting for Z will induce additional correlation between E and D Note that the following two properties hold for every sufficient adjustment set S e S does not contain any variable that lies on a causal path between exposure and outcome indermediate This implies that it is never appropriate to adjust for a variable that is a successor of the exposure e S contains all variables that are direct parents of both exposure and outcome 4 2 Finding minimal sufficient adjustment sets Whenever you draw a causal model using the button draw DAG or make changes to it DAGitty will calculate all minimal sufficient adjustment
12. s are ignored by DAGitty By convention the variable in the first line is the exposure and the variable in the second line is the outcome of your model Variable names must not contain spaces or colons please use dashes or underscores instead i e write fitness_level instead of fitness level The list of connection consists of several lines each starting with a start variable name followed by one or more other target variables that the start variable is connected to Figure 3 contains a working example of a textual model description When you modify a model within DAGitty the vertex labels will be augmented by additional information to help DAGitty remember the layout of the vertices and for other purposes see rightmost column in Figure 3 2 2 Loading a model into DAGitty To load a textually defined model into DAGitty simply copy amp paste the variable list into the vertex labels text field and the list of connections into the adjacency list Then click on draw DAG DAGitty will now generate a preliminary graphical layout for your model on the canvas which may or may not be aesthetically pleasing but can be freely modified 2 3 Modifying the graphical layout of a model To layout the vertices and edges of your model more clearly than DAGitty did simply drag the vertices with your mouse on the canvas You will notice that DAGitty modifies the list of vertices in the vertex labels text field on the fly and augme
13. uch that adjustment stratification or selection e g by restriction or matching will minimize bias when estimating the causal effect of the exposure on the outcome You can read more about controlling bias and counfounding in Pearl s textbook chapter 3 3 and epilogue 6 Moreover Shrier and Platt give a nice step by step tutorial on how to test if a set of covariates is a sufficient adjustment set Briefly a sufficient adjustment set S blocks all non causal paths between exposure and outcome but leaves open all causal paths i e chains of the form e gt z gt gt x gt 0 A path p is blocked by a set Z if at least one of the following properties holds 6 e The path p contains a chain z gt m gt y or a fork x m gt y such that m is in Z e The path p contains a collider x c y such that c is not in Z and furthermore Z does not contain any successor of c in the graph A path p is called open if it contains no collider and at least one fork and closed if it contains at least one collider Every non causal path is either open or closed As proved by Lauritzen et al 5 see also Tian et al II it suffices to restrict our attention to the part of the model that consists of exposure outcome and their ancestors for identifying sufficient adjustment sets This is indicated by DAGitty by coloring irrelevant nodes in gray The relevant nodes are colored according to which node they are ancestors of exposu
14. we control for smoking i e put smokers and non smokers in two different groups we would probably no longer find a correlation between carrying matches and lung cancer In general any set of variables in a causal model that blocks all confounded i e non causal effects between an exposition and an outcome but does not affect the causal effects is called a sufficient adjustment set If the causal model is accurate then adjustment stratification or selection e g by restriction or matching for this set of variables in an epidemiological study will minimize bias when estimating the effect of exposition on outcome in an epidemiological study Adjustment sets will be explained in more detail in Section 4 The purpose of DAGitty is to aid epidemiological study design through the identification of suitable small sufficient adjustment sets in complex causal models There are two ways to run DAGitty either from the internet or from your own computer 1 2 Running DAGitty online To run DAGitty online simply open its URL in your favourite Browser 1Calling bayesian networks DAGs is of course highly confusing to computer scientists and mathematicians for whom a DAG is simply an abstract graph without specific semantics attached to it http www tcs uni luebeck de software dagitty DAGitty should run in every modern Browser If it doesn t please send me an E Mail so I can fix the problem see contact information at the end of t
Download Pdf Manuals
Related Search