# Does this rdists package provide actual (predicted) response time for each observation?

With the help of other packages, I've estimated Drift Diffusion model parameters of my data. Now, I want to estimate the predicted (or actual) response times for each observation with the help of estimated parameters. By going through the manual I got a sense that ddiffusion function estimates (predicts) the actual response times, basically I used this code for my own purposes:

dd1<-ddiffusion(data$rt , data$resp,a=2.16,v=1.12,t0=0.36,z=0.51).


Does this function gives me actual (predicted) response time for each observation?

Not very familiar with this package, but per manual (bold and italics added):

ddiffusion gives the density, pdiffusion gives the distribution function, qdiffusion gives the quantile function (i.e., predicted RTs), and rdiffusion generates random response times and decisions (returning a data.frame with columns rt (numeric) and response (factor)).

So, it looks like if you want predicted reaction times, you should use the qdiffusion function (but note that it is a quantile function).

• mfloren, thanks for the contribution. When I run qdiffusion returns, for instance, 2.27 when the observed response time is 0.62. It doesn't make sense, as predicted RT cannot be higher than the observed RT. Jul 22, 2017 at 16:16
• @Samir The qdiffusion gives quantiles of the predicted values based off of "minimizing the absolute difference between desired probability and the value returned from pdiffusion using optimize." Have you gotten a chance to look through the examples in the documentation? That is where all of this information is coming from. Jul 22, 2017 at 16:40
• Yes, I am constantly reading the manual. Apparently, I have to spend more time on it. Thanks Jul 22, 2017 at 16:41
• @Samir You're welcome. I am not an expert of this package by any means, and perhaps others can give a more satisfactory answer. I'll update mine to include that. Jul 22, 2017 at 16:42

The functions estimate the parameters of the distribution. If I understand you correctly, you want to make individual predictions - these require a predictor variable in the sense of a regression model. (sorry for the late response, I just stumbled across the question)

• Thanks - appreciate your contribution! Oct 27, 2019 at 2:41