Define constraints for combinations of analysis paths in the multiverse

Users can specify conditions for which a particular analysis path may or may not be valid using either the %when% operator or the branch_assert function.

Arguments

Logical

predicates defined in terms of the parameters and their options in the multiverse.

Details

A user can specify multiple different analysis options at each step of the analysis process using branch(). However, it is possible that the values of certain parameters might be conditional on the values of certain other parameters.

The conditional or the "implies" relationship between two statements, A and A has the meaning, "if A is true, then B is also true." We evaluate this relationship using classical logic: \(A \Longrightarrow B\) is an abbreviation for \(\neg A \lor B\)

See also

vignette("conditions")

Examples

# \donttest{
M.1 <- multiverse()

# There are different ways to specifying conditions
# One way is to use the %when% operator
# the %when% operator can be specified after the option name
inside(M.1, {
    df <- data.frame (x = 1:10 ) %>%
        mutate( y = branch( values_y, TRUE, FALSE )) %>%
        mutate( z = branch(values_z,
            "constant" ~ 5,
            "linear" ~ x + 1,
            "sum" %when% (values_y == TRUE) ~ (x + y)
        ))
})

# or it can be specified after the expression for computing the option value
inside(M.1, {
    df <- data.frame (x = 1:10 ) %>%
        mutate( y = branch( values_y, TRUE, FALSE )) %>%
        mutate( z = branch(values_z,
            "constant" ~ 5,
            "linear" ~ x + 1,
            "sum" ~ (x + y) %when% (values_y == TRUE)
        ))
})

# an advantage of the '%when' operator is that it can also be used when the
# option names are not specified for branches.
# when option names are not specified for branches, option names are assigned to
# the branches. For character, logical or numeric expressions, option names are of the
# same type (i.e. character, logical or numeric expressions respectively)
# For expressions of type symbol or call, options names are characters strings
# containing the expression.
# see the next two examples:
M.2 <- multiverse()

inside(M.2, {
 df <- data.frame (x = 1:10 ) %>%
   mutate( y = branch( values_y, TRUE, FALSE )) %>%
   mutate( z = branch(values_z,
          5,
          x + 1,
          (x + y) %when% (values_y == TRUE)
   ))
})

M.3 <- multiverse()
inside(M.3, {
 df <- data.frame (x = 1:10 ) %>%
   filter( branch( values_x,
       TRUE,
       x > 2 | x < 6
   )) %>%
   mutate( z = branch(values_z,
          5,
          x + 1,
          (x^2) %when% (values_x == 'x > 2 | x < 6')
   ))
})

# or it can be specified after the expression for computing the option value
M.4 <- multiverse()
inside(M.4, {
    df <- data.frame (x = 1:10 ) %>%
        mutate( y = branch( values_y, TRUE, FALSE )) %>%
        mutate( z = branch(values_z,
            "constant" ~ 5,
            "linear" ~ x + 1,
            "sum" ~ x + y
        )) %>%
        branch_assert( values_z != "sum" | values_y == TRUE )
})
# }