Assessing the ability of matching to address attrition bias in a randomized experiment using data from the rand health insurance experiment. Selection bias is the bias introduced by the selection of individuals, groups or data for analysis in such a way that proper randomization is not achieved, thereby ensuring that the sample obtained is not representative of the population intended to be analyzed. A systematic error caused by the selective occurrence and biased handling of protocol deviations and losses to followup, which may lead to results. As opposed to parametric approaches to correcting for sample selection bias. A tale of two surveys the randomized trial literature has helped to renew the fields of microeconometric policy evaluation by emphasizing identification issues raised by endogenous program participation. Selectivity bias in panel data hausman and wise 1979 proposed a method to deal with panel attrition that uses a probability model of attrition in conjunction with a traditional random effect model of individual response. If the test for invariance is nonsignificant, then the two matrices are assumed to be equivalent, with no apparent attrition bias. Using data from a split panel with nearly identical questionnaires enabled us to gauge the initial attrition bias for variables of interest by comparing the distribution of a variable between a. To fix ideas consider a household panel consisting of i. Apr 27, 2004 any hypothesis in longitudinal studies may be affected by attrition and poor response rates. Different rates of loss to followup in the exposure groups, or losses of different types of participants, whether at similar or different frequencies, may change the characteristics of the groups, irrespective of the exposure or intervention.
Learn vocabulary, terms, and more with flashcards, games, and other study tools. Attrition threatens the internal validity of cohort studies. Does attrition during followup of a population cohort study. Using multiple imputation to deal with missing data and attrition in longitudinal studies with repeated measures of patientreported outcomes karin biering,1 niels henrik hjollund,2,3 morten frydenberg4 1danish ramazzini centre, department of occupational medicine university research clinic, hospital west jutland, herning, denmark. Aug 24, 2015 if you would like to learn more about treatment effects in stata, there is an entire manual devoted to the treatmenteffects features in stata 14. We simulated a cohort of 300 subjects using 500 computer replications to determine whether regression imputation, individual weighting, or multiple. We include examples of both a fully bayesian approach for analyzing the concatenated panel and refreshment data, and a multiple imputation approach for analyzing only the original panel. Correcting for attrition bias statalist the stata forum. Testing and adjusting for attrition in household panel data. Attrition occurs when study participants who were assigned to the treatment and control conditions do not provide outcome data and thus do not contribute to the estimation of the treatment effects.
However, this is not the case when the outcome variable is defined as the total health care expenditures. Eric ed517992 biases in estimating treatment effects due. We provide a statistical framework for conducting tests for attrition bias that draws a sharp distinction between selection on unobservables and on observables and that shows that weighted least squares can generate consistent parameter estimates when selection is based on observables, even when they are endogenous. The main purpose of this study is to elaborate a model of the relationships between attrition and effect estimates and to use that model to guide monte carlo simulations that examine the sources and magnitude of attrition bias under various assumptions for randomized experiments and cluster randomized experiments. Our analysis used data from a sevenwave webbased panel survey that had been supplemented by three crosssectional surveys. Attrition and bias in the mrc cognitive function and.
The stata blog solving missing data problems using inverse. Such attrition may bias the estimated effect of the experiment. Methods to account for attrition in longitudinal data. Social experiments frequently exploit data from administrative records. Attrition bias definition of attrition bias by medical. N households who have been surveyed in two different years t1, 2. We discuss estimating populationaveraged parameters when some of the data are missing. Does attrition during followup of a population cohort. Oct 11, 2016 we discuss estimating populationaveraged parameters when some of the data are missing.
Eric ed517992 biases in estimating treatment effects. Note that one can use stata to estimate both types of models. To install an ado file, type ssc install xxx where xxx should be replaced with the name of. In the case of attrition bias, this method is now the industry standard.
It is a weighted gmm estimator, that solves in the justidentified. It is sometimes referred to as the selection effect. Both nonparticipation and attrition may introduce selection bias if the reason for is related to the outcome of interest. I started with a sample size of 88 individuals and for the next tow years 200506 and 200607 30 individuals dropped out and for the last period 200708 one individual dropped out. Correction for attrition bias classical approaches. Nonrandom sample selection may render estimated treatment effects biased. The statistical package stata will be used to illustrate the methods during. If you would like to learn more about treatment effects in stata, there is an entire manual devoted to the treatmenteffects features in stata 14. Jan, 2005 the elegant study by tierney and stewart 11 for the first time directly assesses the impact of attrition bias. Attrition bias is one of the major threats to multiwave studies, and it can bias the sample in two ways. Pdf attrition, selection bias and censored regressions.
Now im trying to use inverse probability weighted ipw estimators to deal with the attrition bias. Measurement and attrition issues have perhaps received less attention. We provide a statistical framework for conducting tests for attrition bias that draws a sharp distinction between selection on unobservables and on observables and that shows that weighted least squares can generate consistent parameter estimates when selection. Attrition and bias in the mrc cognitive function and ageing. Selection bias correction based on the multinomial. Vella, francis 1998, estimating models with sample selection bias. The major advantage of this design is that we can measure bias with reference to a large quantity of substantive variables of interest. The module is made available under terms of the gpl v3. Attrition or exclusion bias results from participants in a study not being included in the results when there are systematic differences between the people that are included, and the people that are not. However, the ability of these methods to correct for attrition bias has not been tested. We describe the patterns of nonresponse and examine if attrition affects the econometric analysis of medical labour market outcomes using the estimation of physician earnings equations as a case study.
Stata module for estimating lee 2009 treatment effect bounds. The mrc cognitive function and ageing study mrc cfas is a population based longitudinal study in five centres with identical methodology in england and wales each recruiting approximately 2,500 individuals. They do not feel that they are in any way responsible for attrition. Although differential dropout can bias results, it does not always do so. Attrition bias is a kind of selection bias caused by attrition loss of participants, discounting trial subjectstests that did not run to completion. Assessing the ability of matching to address attrition bias. Dear all, i have a problem of attrition bias in my panel data the years 200408.
Luis huesca mario camberos centro conacytde investigacion en alimentacion y desarrollo, a. The main advantage of paradata is that it is collected as a by product of the survey process. Finally, without conditioning on year 1 rank, matching cannot mitigate attrition bias for either outcome variable. Differential dropout and bias in randomised controlled. Its implementation shows the usefulness of structures and pointers. This estimator estimates a vector of parameters, defined by a set of moment conditions. The phrase selection bias most often refers to the distortion of a statistical. Attrition occurs when participants leave during a study. In this paper we present the results of a study of attrition and its potential bias in one of the most wellknown panel data sets, the michigan panel study of income dynamics psid. Risk of attrition in a longitudinal study of skin cancer. Dropout in randomised controlled trials is common and threatens the validity of results, as completers may differ from people who drop out. Similarly, equal dropout may or may not lead to biased results. The distortion in the inference on the conditional regression function, or some component thereof, induced by such a mechanism is referred to as attrition or selection bias. This paper investigates the nature and consequences of sample attrition in a unique longitudinal survey of medical doctors.
It is very common in experimental studies in education as illustrated, for instance, in a metaanalysis studying. A number of approaches to handle attrition exist, and their use depends on the assumptions made about the origins and causes of the missing data problem. Survival analysis showed that risk of attrition in a longitudinal study was increased among participants diagnosed with skin cancer, whereas the converse was the case under a logistic approach. Here is a familiar scenario for those running field experiments. The application of a split panel design allows us to estimate the bias due to panel attrition by comparing the panel waves to the crosssectional surveys at the item level. Any hypothesis in longitudinal studies may be affected by attrition and poor response rates. This module should be installed from within stata by typing ssc inst firthlogit. Assessing the ability of matching to address attrition. Stata usersgroupmeeting mexico discussion of userwritten stataprograms selection bias correction based on the multinomial logit. In particular, we show how to use gmm to estimate populationaveraged parameters for a probit model when the process that causes some of the data to be missing is a function of observable covariates and a random process that is independent of the outcome. The psid has suffered a large volume of attrition since it began in 1968almost 50. Epidemiologists use various imputation and weighting methods to limit bias due to attrition.
It is closely related to the survivorship bias, where only the subjects that survived a process are included in the analysis or the failure bias, where only the subjects that failed a. Anonymised primary care data, routinely collected in clinical practice and independent of any cohort research study, represents an ideal unselected comparison dataset. If the data are assumed to be 1 see jones 2007 for an extensive list of longitudinal surveys used in applied research on health economics. Sep 24, 2017 in field experiments with survey or biomarker data collection, there will always be loss to followup and without some assumptions about the form of that attrition, it may be impossible to make any causal inferences about intention to treat or average treatment effects. Stata module to calculate bias reduction in logistic regression, statistical software components s456948, boston college department of economics, revised 25 jul 2015. Differing dropout rates between treatment arms is sometimes called differential dropout or attrition.
Using paradata to predict and correct for panel attrition citeseerx. First, attrition bias can affect the external validity of the study. Youre conducting a study with a treatment and a comparison arm and measuring your main outcomes with surveys andor biomarker data collection, meaning that you need to contact the subjects unlike, say, using administrative data tied to their national identity numbers preferably in person. Also, they feel that there is a correlation among the age of employees who leave and the reasons they leave. When an experimental participant moves across state lines, his entries in the data system of his state of origin consist entirely of zeros. It is a weighted gmm estimator, that solves in the.
However, most administrative data systems are designed to track earnings or benefit payments among residents within a single state. Person stay in study until end differs from those who drop out during study cohort study. If refreshment samples are available, something can be done. Logistic modeling has commonly been used to identify attrition risk factors in longitudinal studies. As a result, the remaining sample is not generalizable to. Using paradata to predict and correct for panel attrition. The psid has suffered a large volume of attrition since it began in 1968almost 50 percent of initial sample members had attrited by 1989.
Montecarlo comparisons delta working paper 200420, download. Attrition is a potential source of bias in cohort studies. This paper aims to identify potential biases in the twoyear followup interviews. Attrition is estimated from runs of zeros at the end of the sample period.
Furthermore, this design enabled us to analyze in detail the effectiveness of propensity score weights. If attrition is taken into account, usually mar selection on observables is assumed. Special considerations with unbalanced panels will be discussed, including how to test for sample selection and attrition bias. View or download all content the institution has subscribed to.
People included in study are unrepresentative due to sampling or selection factorsnon participation. This split panel design allowed us to assess the magnitude of attrition bias for a large number of substantive variables. This module should be installed from within stata by typing ssc install. There are some observations from the baseline years missing in. Full text using multiple imputation to deal with missing. Stewart and colleagues have coordinated individual patient data metaanalyses of clinical trials in oncology for many years and now present a comparison of the results from their analyses, which invariably followed the intention to.
Participation sipp, shows little indication of bias due to attrition in a model of labor market 1 see for instance the special issue attrition in longitudinal surveys, of the journal of human resources, spring, 1998, vol. The elegant study by tierney and stewart 11 for the first time directly assesses the impact of attrition bias. Testing for attrition bias using the bglw test stata. Floyd university of toronto july 22, 20 this document outlines the nature of simultaneity bias in regression analysis and the standard method for dealing with it, twostage least squares. Panel data methods cemmap centre for microdata methods and. As we shall see, sample selection bias can be viewed as a special case of endogeneity bias, arising when the selection process generates endogeneity in the selected subsample. If some groups of people drop out of the study more frequently than others, the subsequent longitudinal sample no longer resembles the original sample in the study. To estimate the attrition arising from interstate mobility and provide bounds on the effect of the experiment. The center management feel that attrition is a function of age only. See my previous post on excel chitest to see what my results showed regarding this. A structural equation modeling program, such as lisrel or amos, can be used to test whether the two correlation matrices are invariant, that is, the same. Although attrition may be inevitable in cohort studies of older people, there is little empirical evidence as to whether bias due to such attrition is also inevitable.
1427 173 108 1036 1013 401 1075 1202 125 271 768 5 1476 472 1010 381 1360 1594 1019 150 1606 772 1182 519 60 1233 263 20 160 765 961 262 888 548 1125 1489 67 333 1494 191 138 940 706 879 20