Dear all,
I intend to use a matching procedure on a project where, because of the nature of my cases, weights of control units should be all kept at 1. Reading the manuals and articles I realized that the way to do it was using CEM and applying the function k2k.However, I couldn't get it to work as described. When I run the command matchit(x, ..., k2k=TRUE), it returns a matrix that, when extracted into a data-frame, still contains weights and a different number of treated and control units - even though no error messages are shown. This happens both to my data and when running the lalonde demo script. The result of the matching is exactly the same as that obtained when k2k=FALSE. I tried to go back to the "cem" package and run the demo script there on LL with k2k - using first the command cem(x,... k2k=TRUE) and also doing the matching first and applying k2k(cem.match,...) afterwards. In both cases there's the error message "In min(x, na.rm = TRUE) : no non-missing arguments to min; returning Inf", and the resulting matrix is still exactly the same as that obtained without setting k2k to TRUE.
Thank you in advance for the help.
Best regards,
Bruno Castanho e Silva
Dear All,
I am brand new with R and propensity score. I am using SAS. I do not know
whether I can use matchit for my matching or not. I want to match stocks
listed in NASDAQ with those listed in NYSE.
My independent variables are log (market value of stock at the end of each
month), log (price at the end of each month), industry dummy, and year
dummy. Actually, there are several more variables. However, I only show a
few of them.
At the first step, I run a logit for all years and all industries to get
the propensity score.
At the second step, each month from 1980-2013, I match NASDAQ stocks with
NYSE stocks within the same industry (15 industries) using nearest
neighbor.
Do you have any suggestions?
Thank you so much for your help,