Use named$-$based selection to reorder variables as name, x$3,$ x$2,$ and x$1,$ and use criterion$-$based selection to select only the observations where x$1$ is greater than $2$ and not equal to x$3($from the full dataset$).$ Then add a synthetic variable that expresses the means of the numeric variables. Then use timestep offset to add a synthetic variable that expresses the difference between the means of adjacent observations.

You may want to use these function$($s$)$:

rowMeans$()$

nrow$()$