Skip to contents

Crosswalk an effect estimate

do_crosswalk.Rd

Take a published effect estimate (e.g., the difference in Mini Mental State Exam score comparing APOE-ε4 carriers to non-carriers) and translate that effect estimate to an alternate scale (e.g., Montreal Cognitive Assessment). The translation uses a crosswalk estimated via crosswalk() in data where both measures are available.

Usage

do_crosswalk(
  object,
  est_mean = NULL,
  est_se = NULL,
  est_ci = NULL,
  est_pval = NULL,
  est_alpha = 0.05,
  est_indep = NULL,
  est_outcome = NULL,
  alpha = 0.05
)

Arguments

object

An object of class cogxwalkr or the result of est_cw_coef()

est_mean

Point estimate (beta) to be crosswalked to the alternative outcome measure

est_se

The standard error of est_mean

est_ci

The lower (1-alpha)% confidence interval of est_mean

est_pval

The p-value corresponding to est_mean

est_alpha

The alpha level for the confidence interval (if est_se is provided) or the alpha level that will be used to back-calculate the standard error from est_ci. Defaults to 0.05.

est_indep

The independent variable to which est_mean applies

est_outcome

The outcome measure in the original study (e.g., "MOCA", "MMSE")

alpha

The alpha level for the confidence interval of the crosswalked estimate. Defaults to 0.05.

Details

Parameters prefixed with est_ refer to a summary estimate for which the user lacks access to the underlying data but wishes to translate the estimate to another cognitive measure's scale. The user must supply est_mean and one of est_se, est_ci, or est_pval. do_crosswalk() will back-calculate the standard error if necessary, as follows:

  • est_ci : (confidence interval width) / 2 / (critical value), where "critical value" refers to the Z-value of the standard normal distribution assuming a two-sided est_alpha

  • est_pval : est_mean / (critical value), where "critical value" in this case is calculated assuming a two-sided p-value

As in the Cochrane Handbook summary of these calculations, the function assumes that statistical estimates for difference measures were calculated using the standard normal distribution rather than a t-distribution.