I have the following data.table:
Dates variable value
1: 1985/01/02 SLB UN Equity 4.5192
2: 1985/01/03 SLB UN Equity 4.3111
3: 1985/01/04 SLB UN Equity 4.2963
4: 1985/01/07 SLB UN Equity 4.2368
5: 1985/01/08 SLB UN Equity 4.1327
---
I would like to perform "n"-number of "shifts" to value-column, and then calculate the return by dividing value-column by each "shifted" column. I have managed the first part using:
dt[, paste("px.min.", 1:2, sep = "", "d") := shift(value, 1:2), by = variable][]
with the result being:
Dates variable value px.min.1d px.min.2d
1: 1985/01/02 SLB UN Equity 4.5192 NA NA
2: 1985/01/03 SLB UN Equity 4.3111 4.5192 NA
3: 1985/01/04 SLB UN Equity 4.2963 4.3111 4.5192
4: 1985/01/07 SLB UN Equity 4.2368 4.2963 4.3111
5: 1985/01/08 SLB UN Equity 4.1327 4.2368 4.2963
---
The 2nd part is causing the problem.
dt.px.melt[, paste("px.min.", 1:2, sep = "", "d") := value/shift(value, 1:2), by = variable][ ]
produces the error:
"Error in value/shift(value, 1:2) : non-numeric argument to binary operator"
You can use
dt[, paste("px.min.", 1:2, sep = "", "d") :=
lapply(shift(value, 1:2), function(x) value/x), by = variable][ ]
The point being that you need to conduct the division operation column by column, which lapply will do for you