Hello all, I have a brief IRT related question regarding the ltm package. I am am using tpm which is part of the ltm package. I am constraining the item parameters for this package since they have been estimated earlier. I notice that the difficulty parameter changes, despite the fact that it is supposed to be constrained. The a and c parameters do not change. Here is what I am doing (in principle): #example values for the constraints difficulties <- c(-0.5, 1, 0.1, 1.25) discrimination <-c(0.6, 1.1, .5, .3) items <- rep(1:4, 3) parms2 <- rep(1:3, each = 4) vals <- c(rep(.25, 4),difficulties, discrimination) constr2 <- cbind(items, parms2, vals) constr2 #note that dataframe LSAT comes with the ltm package tpm(LSAT, type = "latent.trait", constraint = constr2) Note that the difficulty parameter changes. Do any of you have any suggestions why that is the case? Does this influence the estimate of the other item parameter (the ones I am trying to estimate?) Your thoughts & help would be greatly appreciated, Alexander [[alternative HTML version deleted]]