Cem

cem@lists.gking.harvard.edu

152 discussions

Possibly unexpected behavior in cem and MatchIt

by Alberto Palomar

Hello, Posted the below to the MatchIt mailing list and was advised there might also be interest here. Have modified it slightly with some debugging information on an error that comes out of cem's att() function when called with the treatment variable as a factor: ----- Hello, Firstly thank you for this very useful package. I came to it via the CEM package in the course of trying to explore how some versions of PSM would compare on some clinical data we are analyzing to retrospectively make an estimate of how effective a certain therapy is for moderately severe hospitalized Covid-19 cases. While doing this there seemed to be some unexpected behavior with respect to whether the treatment variable is included as an integer or as a factor, which I do not think I understand after reading the package documentation, vignettes, and the explanation of weights (at https://docs.google.com/document/d/1xQwyLt_6EXdNpA685LjmhjO20y5pZDZYwe2qeNo…). Briefly, when the treatment variable is a factor, the output of matchit with the cem method appears to use the opposite convention for assigning weights [i.e. *un*treated weights are 1 and treated weights are presumably (m_C/m_T)*(m^s_T/m^s_C) ] from what comes out of cem directly (regardless of whether the treatment variable is integer or factor in the cem input data). So then after applying match.data() to the matchit outputs, the results of subsequent regressions (have tested lm and glm) with the corresponding weights also differ depending on whether the treatment variable is coded as integer or as a factor. Is this an expected behavior? Here is a somewhat minimal example using the LaLonde data, with apologies for its inelegance: #### library(cem) library(MatchIt) data("LeLonde") #set up a data frame with treatment as integer (Le) and another with #treatment.f as factor (Le.f) Le <- data.frame(na.omit(LeLonde)) Le.f <- Le Le.f$treated <- as.factor(Le.f$treated) colnames(Le.f)[which(colnames(Le.f) == 'treated')] <- 'treated.f' #for simplicity will match only on age, re74, and q1; here make lists #of the other variables to drop in the cem commands LeLonde_vars_to_match <- c("age", "re74", "q1") LeLonde_vars_to_drop <- setdiff(names(Le), LeLonde_vars_to_match) LeLonde_vars_to_drop.f <- setdiff(names(Le.f), LeLonde_vars_to_match) #use cem directly to match between treated/untreated on age, re74, and #q1; first with treated as integer, second with treated.f as factor mat.cem <- cem(treatment = "treated", data = Le, drop = LeLonde_vars_to_drop, eval.imbalance = TRUE, keep.all = TRUE) mat.f.cem <- cem(treatment = "treated.f", data = Le.f, drop = LeLonde_vars_to_drop.f, eval.imbalance = TRUE, keep.all = TRUE) identical(mat.cem$w, mat.f.cem$w) #is true, weights from cem do not #depend on whether treatment variable is integer or factor #use matchit with method cem to match between treated/untreated on #age, re74, and q1; first with treated as integer, second with #treated.f as factor mat.mi <- matchit(treated ~ age + re74 + q1, Le, method = "cem") mat.f.mi <- matchit(treated.f ~ age + re74 + q1, Le.f, method = "cem") identical(mat.mi$weights, mat.f.mi$weights) #is false, weights from #matchit with method cem do depend on whether treatment variable is #integer or factor, seemingly by different choice of whether control or #treated weights are set to 1, as the same data entries are selected by #the match, as suggested by plot(match.data(mat.mi)$re78, #match.data(mat.f.mi)$re78) being linear with slope 1 identical(mat.cem$w, mat.mi$weights) #is true, indicating that the #convention from cem coincides with using matchit with method cem when #the treatment variable is integer #generate matched datasets, first with treated as integer, second with #treated.f as factor matched.data.mi <- match.data(mat.mi) matched.data.f.mi <- match.data(mat.f.mi) #compute the linear models for re78 ~ treated, first with treated as #integer, second with treated.f as factor mod.mi <- lm(re78 ~ treated, matched.data.mi, weights = matched.data.mi$weights) mod.f.mi <- lm(re78 ~ treated.f, matched.data.f.mi, weights = matched.data.f.mi$weights) identical(mod.mi$coefficients["treated"], mod.f.mi$coefficients["treated.f1"]) #is false, the estimates depend #on whether the treatment variable is integer or factor mod.mi.g <- glm(re78 ~ treated, data = matched.data.mi, weights = matched.data.mi$weights) mod.f.mi.g <- glm(re78 ~ treated.f, data = matched.data.f.mi, weights = matched.data.f.mi$weights) identical(mod.mi.g$coefficients["treated"], mod.f.mi.g$coefficients["treated.f1"]) #is false, the estimates depend #on whether the treatment variable is integer or factor #### There is a separate peculiarity when using cem directly with the treatment variable as a factor, where the att() command fails with error messages saying that treated.f is not a factor, for instance after running the above code: #### est.f.cem <- att(mat.f.cem, re78 ~ treated.f, data = Le.f) #Error: variable 'treated.f' was fitted with type "factor" but type #"numeric" was supplied #In addition: Warning message: #In model.frame.default(Terms, newdata, na.action = #na.action, xlev = object$xlevels) : # variable 'treated.f' is not a factor #### Debugging this it seems that in line 292 of att.R in the github repo, a temporary data variable's treatment column is assigned 0, which makes it a numeric type, and then this goes to the predict() function along with the cem object which still expects the original factor type, which I think is the root of the discrepancy. Maybe adding a line there checking if the treatment variable is a factor, and if so casting the temporary data's treatment column to a factor if that is so, would be appropriate? In the short term I think a reasonable solution is just not to cast the treatment variable as a factor in matchit when using cem or when using cem/att - am sorry if this was advised in the documentation somewhere and I missed it - but it might be worth considering if matchit should be able to tolerate having the treatment variable as a factor as well (or raise a runtime alert about possibly different output or potential errors). Thank you, Alberto

3 years, 11 months

Automating progressive coarsening

by Cephas N Ablakwa

Hi Stefano, I sent you a couple of emails last week to your iacus account. Did you get them? Was I suppose to send them to this address instead?

3 years, 11 months

Automating progressive coarsening

by Cephas N Ablakwa

Hi, My name is Cephas Ablakwa, and I am a PH.D. student studying in Program Planning and Evaluation in education. I am working on my dissertation, hoping to defend in June 2020. While using the CEM package in R, I ran into an issue that I hope you can help me with. When automating a progressive coarsening procedure using the relax.cem function, I got the following error: Error1:n : argument of length 0. After googling the issue extensively, I found an answer to a similar error (but with a different function) that suggested there might be a matrix with "one dimension of length 1, without drop=FALSE, R converts a matrix to a vector." The solution offered to this similar problem was lengthy, complex, and specific to the issue. Given my level of experience with R, I do not know how to extrapolate this solution to my issue. In my field and at my university, we mainly use SPSS. I have taken up R mainly to use your CEM package and explore causal inference and coarsened exact matching in my work. I have tried multiple solutions, including restructuring the class of my variables to resemble the class structure of the LeLonde dataset used in the example. Find attached is a summary of the initial steps I took to arrive at the error. I could provide more information if you need it. Any help you can offer will be greatly appreciated. I hope to hear from you soon. Cephas Ablakwa Tennessee Tech University College of Education, PH.D. Candidate |TJ Farr Room 105 cnablakwa21(a)students.tntech.edu [1505503380663_PastedImage]

3 years, 11 months

How to apply CEM to a multi-category treatment

by Alexi Chacon

I am currently trying to use CEM with a multi-category treatment in R. When I try to run a regression between the outcome and the treatments, it does not treat my treatment variable as a multi-level variable. Is it possible to have all treatments in one variable using CEM or does each treatment need to be regressed in a separate model? If it is possible, how is the code written? Best, Alexi

3 years, 11 months

Question regarding indexing of strata options in imbspace package

by Aaron Wells

To whom it may concern: I have successfully been able to execute the imbspace package. I am struggling with identify the coarsenings that apply to a given stratification option. For example, I am interested in knowing the cutpoints used to create #3 below. However, upon running the $coars[[3]] object for this row, the result does not match the $space[3, ] result. I must be missing something obvious as the rows should be indexed identically across objects within a given package. To this end, and apologies for the simple request, but can you help me match the coarsenings to the stratification option. Thank you. Aaron > imbsp_bgnyr$space G0 G1 PercG0 PercG1 Relaxed ML1 var 1 8706 1865 96.29466 91.91720 RTM_MM(3), age(3), PHC_predictive_model_score(3) 0.2554752 PHC_predictive_model_score 2 8596 1867 95.07798 92.01577 RTM_MM(3), PHC_predictive_model_score(3), rtm_rpt_gap_mxrxpmpm(3) 0.2599375 rtm_rpt_gap_mxrxpmpm 3 8614 1866 95.27707 91.96649 RTM_MM(3), PHC_predictive_model_score(3), rtm_pmpm(3) 0.2600584 rtm_pmpm [cid:image002.jpg@01D62C6F.56EB9500] Aaron Wells, PhD Vice President, Outcomes PopHealthCare C: 615-417-5476 E: aaron.wells(a)pophealthcare.com<mailto:aaron.wells@pophealthcare.com> www.pophealthcare.com<http://www.pophealthcare.com/> [C9198E52] IMPORTANT NOTICE: This e-mail message is intended to be received only by persons entitled to receive the confidential information it may contain. E-mail messages to clients of PopHealthCare may contain information that is confidential and legally privileged. Please do not read, copy, forward, or store this message unless you are an intended recipient of it. If you have received this message in error, please forward it to the sender and delete it completely from your computer system.

3 years, 11 months

balance

by fang zhang

Hi All, We often find that treating meaningful binary variables (e.g gender) as continuous variable (autocuts Sturges) will result better balance in demographic table 1 in terms of standardized differences (of course less sample size, usually not much loss for treatment arm). Should we use the Sturges autocuts rather than more meaningful binary variables? Thanks, Fang

4 years, 8 months

Very different ATT estimates using method = 'linear-RE'

by Juan Tellez

Hello, I'm working in R, and trying to understand the decision, post-matching, between using model = 'linear' and model = 'linear-RE' in the specification of att(), like this: ## data(LL) todrop <- c("treated","re78") # cem match: automatic bin choice mat <- cem(treatment="treated", data=LL, drop="re78") mat # models att(mat, re78 ~ treated, data=LL, model = 'linear') att(mat, re78 ~ treated, data=LL, model = 'linear-RE') The SATT is a bit different across models, but what's really striking is how relatively tiny the standard errors are when model = 'linear-RE'. Is there something I could read on the motivation behind 'linear-RE', contexts where it might be appropriate, etc.? Thank you, Juan Tellez -- Juan Fernando Tellez http://juanftellez.com Department of Political Science Duke University

5 years

Error with Imbalance Command in Stata 15

by Cisney-Booth, Ann M.

Good morning, I am using CEM for the first time and going through the Blackwell, Iacus, King, and Porro (2009) article in the Stata Journal by following the commands. However, I cannot get the imbalance command to work. I receive the following error message: . imbalance age education black nodegree re74, treatment(treated) command imbalance is unrecognized r(199); Has there been an update to the Stata commands since this article was published? Any guidance would be greatly appreciated. Thank you, Ann ________________________ Ann Cisney-Booth, Ph.D. Senior Administrator Strategic Data Project Fellow Research and Evaluation 445 W. Amelia St.; Orlando, FL 32801 (407) 317-3200 ext. 200-2545 www.ocps.net<http://www.ocps.net/> ---------------------------------------------------------------------- The information contained in this e-mail message is intended solely for the recipient(s) and may contain privileged information. Tampering with or altering the contents of this message is prohibited. This information is the same as any written document and may be subject to all rules governing public information according to Florida Statutes. Any message that falls under Chapter 119 shall not be altered in a manner that misrepresents the activities of Orange County Public Schools. [References: Florida State Constitution I.24, Florida State Statutes Chapter 119, and OCPS Management Directive A-9.] If you have received this message in error, or are not the named recipient notify the sender and delete this message from your computer.

5 years

L statistics before matching in SPSS

by Ondřej Hora

Hello all, please can you advise me how can I get balancing statistics for unmatched file in SPSS (before matching) so I can compare change in L statistics before and after matching. I cannot find the syntax for this. Thank you very much in advance, Yours sincerely, Mgr. Ondřej Hora, Ph.D. Odborný asistent/Assistant professor [image: logo Masarykova univerzita Fakulta sociálních studii] *Masarykova univerzita | Fakulta sociálních studií* Katedra sociální politiky a sociální práce A: Joštova 10 | 602 00 Brno T: +420 549 493 892 <+420549493892> E: hora(a)apps.fss.muni.cz | W: www.fss.muni.cz

5 years, 2 months

k-to-k Matching using CEM in R

by Fowler, Christopher A.

Hello Dr. King, I am using CEM on for a small pilot study and I am new to historical control matching methodologies. I am interested in using k-to-k matching to draw a matched sample from people who have completed a specific clinical program. I ran the following code and the CEM software seemingly performs the k-to-k match: mat3.1 <- cem(treatment = "Intervention", data = PP, drop = todrop3, keep.all = TRUE) matk2k3.1 <- k2k(mat3.1, PP, method = euclidean, 1) matk2k3.1 G0 G1 All 185 15 Matched 15 15 Unmatched 170 0 However, I also get the following warning: "Error in k2k(mat3.1, PP, euclidean, 1) : object 'euclidean' not found" Moreover, I also get: mat3.1$k2k [1] FALSE I was wondering if you had any advice regarding why I am getting this warning and the 'False' indication regarding the match? My concern is that the match may not be completed correctly. Thank you, Chris

5 years, 2 months

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

Cem