info.xml

<?xml version="1.0" encoding="UTF-8"?>

<info>
<author>DataShop@CMU
  <email>datashop-help@lists.andrew.cmu.edu</email>
</author>
<url>https://github.com/LearnSphere/WorkflowComponents/tree/dev/AnalysisPropensity</url>
<date>2018-01-14</date>
<abstract>The <b>Propensity Matching</b> component helps to select matched samples of the original treated and
 control groups with similar covariate distributions. It conducts matching with propensity scores, using R's <a href="https://cran.r-project.org/web/packages/MatchIt/MatchIt.pdf" target="_blank">MatchIt</a> package. 
<p>See <a href="https://pslcdatashop.web.cmu.edu/LearnSphere?workflowId=2447" target="_blank">demo</a> workflow.</p></abstract>
<description>This component helps to select matched samples of the original treated and control groups with similar covariate distributions. It uses propensity scores to conduct matching.</description>

<inputs>
A tab-delimited or comma-separated text file: contains data to be analyzed.
</inputs>

<outputs>
<ul>
<li>The analysis-summary file: a text file containing the matching analysis result</li>
<li>The pdf file: a pdf file containing analysis result plots</li>
<li>The match-data file (not included for Null method): a tab-delimited text file containging the matched data with their matching parameters: distance, weights and subclass.</li>
<li>The file (not included for Null method): a tab-delimited text file containing the original data (matched and un-matched) with an additional column, matched, to indicate if a row is matched or not.</li>
</ul>
</outputs>

<options>
<ul>
<li><b>Treatment:</b> choose the column that indicates if a row is in treatment or not. Values in this column can only be 0 or 1.</li>
<li><b>Covariates:</b> choose one or more columns to be used in creating the distance measure used in the matching.</li>
<li><b>Method:</b> Full (default) or Null. When method = NULL, no matching will occur. The output with no matching can be used to examine balance prior to matching on any of the included covariates.</li>  
<li><b>Distance:</b> only glm is available currently, meaning that propensity scores are estimated with logistic regression using glm().</li>  
<li><b>Include Exact Argument:</b> choose Yes or No.</li>  
<li><b>Exact:</b> choose one or more columns for which variables exact matching should take place.</li>  
<li><b>Include Mahvars Argument:</b> choose Yes or No.</li>  
<li><b>Mahvars:</b> choose one or more columns to specify the variable(s) used to compute the Mahalanobis distance.</li>  
<li><b>Caliper:</b> the width of the caliper to use in matching. Default is set to 0.01.</li>  
<li><b>Columns to Join Match Data with Original:</b> choose one or more columns to use to identify the original data rows that are matched, i.e. these column(s) are used to join matched and original data.</li>  
<li><b>Run t Test for Match Data:</b> choose Yes or No.</li>  
<li><b>Variable for t Test:</b> choose one column to run t-test against the Treatment column within the matched data.</li>  
</ul>

<p><b>Examples in the <a href="https://pslcdatashop.web.cmu.edu/LearnSphere?workflowId=2447" target="_blank">demo</a> workflow</b></p>

<p><b>Initial balance checking:</b>
<ol>
<li>Treatment: choose "Treatment" column.</li>
<li>Covariates: choose Gender, Race, ELLStatus, FreeorReducedLunch, IEPGroup, RITMean and TotalAbsences</li>
<li>Method: choose Null.</li>
<li>Distance: glm</li>
</ol>
</p>

<p><b>Use all covariates in matching:</b>
<ol>
<li>Treatment: choose "Treatment" column.</li>
<li>Covariates: choose Gender, Race, ELLStatus, FreeorReducedLunch, IEPGroup, RITMean and TotalAbsences</li>
<li>Method: choose Full.</li>
<li>Distance: glm</li>
<li>Include Exact Argument: choose Yes.</li>  
<li>Exact: choose GradeCode as exact matching parameters. </li>  
<li>Include Mahvars Argument: choose Yes.</li>  
<li>Mahvars: choose RITMean to compute the Mahalanobis distance.</li>  
<li>Caliper: 0.01.</li>  
<li>Columns to Join Match Data with Original: choose StudentRandomID to identify the original data rows that are matched.</li>  
<li>Run t Test for Match Data: choose Yes.</li>  
<li>Variable for t Test: RITMean, i.e. t-test is run with formula: RITMean ~ Treatment</li>  
</ol>
</p>

<p><b>Optimized model using Race, ELLStatus, RITMean as Covariates in matching:</b>
<ol>
<li>Treatment: choose "Treatment" column.</li>
<li>Covariates: choose Race, ELLStatus and RITMean</li>
<li>Method: choose Full.</li>
<li>Distance: glm</li>
<li>Include Exact Argument: choose Yes.</li>  
<li>Exact: choose GradeCode as exact matching parameters. </li>  
<li>Include Mahvars Argument: choose Yes.</li>  
<li>Mahvars: choose RITMean to compute the Mahalanobis distance.</li>  
<li>Caliper: 0.005.</li>  
<li>Columns to Join Match Data with Original: choose StudentRandomID to identify the original data rows that are matched.</li>  
<li>Run t Test for Match Data: choose Yes.</li>  
<li>Variable for t Test: RITMean, i.e. t-test is run with formula: RITMean ~ Treatment</li>  

</ol>
</p>



</options>

</info>