Exercise

# Split-sample evaluation

In chapter 2, you used the function window() to subset your returns for graphical purposes. In this exercise, you will use the `window()`

to create two samples: an estimation sample and an evaluation sample. This exercise will illustrate how portfolio weights can differ when changing the estimation window.

To remind you, the function `window()`

has argument of `x`

, `start`

, and `end`

. Where `start`

and `end`

are in the format `"YYYY-MM-DD"`

.

The object `returns`

is loaded in your workspace.

Instructions

**100 XP**

- Create the sample
`returns_estim`

by subsetting`returns`

, where the sample begins on January 1st, 1991, and ends on December 31st, 2003. - Create the sample
`returns_eval`

by subsetting`returns`

, where the sample begins on the first day of 2004 and ends on the last day of 2015. - Create a vector of maximum weights equal to 10%, with a length as the number of columns there are in
`returns`

called`max_weights`

. - Create a portfolio with the estimation sample called
`pf_estim`

, where the maximum weight (`reshigh`

) is set to`max_weights`

. - Create a portfolio with the evaluation sample called
`pf_eval`

, where the maximum weight (`reshigh`

) is set to`max_weights`

. - Create a scatter plot of the evaluation portfolio weights versus the estimation portfolio weights (note that you can use
`$pw`

). If portfolio weights are identical, they should be on the 45-degree line.