nv: Normative values generation and management

In visualFields: Statistical Methods for Visual Fields

Normative values generation and management

Description

Functions to generate and handle normative values. Check section Structure of normative values below for details about how to generate functioning normative values

Usage

nvgenerate(
  vf,
  method = "pointwise",
  probs = c(0, 0.005, 0.01, 0.02, 0.05, 0.95, 0.98, 0.99, 0.995, 1),
  name = "",
  perimetry = "static automated perimetry",
  strategy = "",
  size = ""
)

agelm(vf)

tddef(agem)

ghdef(perc = 0.85)

pddef(ghfun = ghdef(0.85))

lutdef(vf, probs, type = "quantile", ...)

gdef(agem, sdtd, sdpd)

lutgdef(g, probs, type = "quantile", ...)

Arguments

vf	visual field data with sensitivity values
method	method to generate normative values, pointwise ('pointwise') or smoothed with 2-dimensional quadratic functions ('smooth')
probs	numeric vector of probabilities with values in [0,1]. The values 0 and 1 must be included
name	name for the normative values, e.g., "SUNY-IU pointwise NVs". Default is blank
perimetry	perimetry used to obtain normative data, e.g., "static automated perimetry" (default)
strategy	psychophysical strategy used to obtain threshold values, e.g., "SITA standard". Default is blank
size	stimulus size, if the same size was used for all visual field locations or empty (default)
agem	age model to construct the function to obtain TD values
perc	the percentile to obtain the ranked TD value as reference for the general height (GH) of the visual field. Default is the 85th percentile, thus 0.85
ghfun	function used for determination of the GH and PD values
type	type of estimation for the weighted quantile values. See wtd.quantile for details. Default is 'quantile'
...	arguments to be passed to or from methods
sdtd	standard deviations obtained for TD values
sdpd	standard deviations obtained for PD values
g	a table with global indices

Value

nvgenerate returns a list with normative values

agelm returns a list with coefficients and a function defining a linear age model

tddef returns a function for the computation of TD values

ghdef returns a function for the computation of the general height

pddef returns a function for the computation of PD values

lutdef returns a look up table and a function for the computation of the probability values for TD and PD

gdef returns a function to compute global indices

lutgdef returns a look up table and a function for the computation of the probability values for global indices

Structure of normative values

This is one of the most complex structures in visualFields. It is necessary to be able to run statistical analyses of visual fields obtained from perimetry and it requires data from healthy eyes for its generation. The normative values are only as good as the data they are generated from. Two common ways to generate full normative values from a dataset of healthy eyes, are provided in the package, depending on the method selected. The first one, method="pointwise", generates normative values directly from pointwise statistics. The second one, method="smooth", uses a 2D quadratic functions to smooth out those pointwise statistics. Variations or improvements can be regenerated by copying the code in those functions and editing it.

• info information regarding normative values. Info is not necessary to carry out statistics, but is useful for the generation of reports. The fields need not be the same as the ones listed here, although these are used in the reports in vfsfa for single field analysis and vfspa for series progression analysis.

  • name name of the normative values

  • perimetry perimetry device for which normative values are intended

  • strategy psychophysical strategy

  • size stimulus size, e.g. Goldmann size III, size V

• agem The normative values' age model. The default methods' generate age linear models with coefficients for each location in locmap in coeff and the function definining the model in model

• sd standard deviations of the sensitivities, s, total deviation (TD) values, td, and pattern deviation (PD) values, pd

• luts Lookup tables to obtain probability levels for TD and PD values.

  • probs probability levels

  • td, pd lookup tables for TD and PD values at each location in locmaps

  • global lookup table for the following global visual field indices

    • ms mean sensitivity (MS) calculated as the unweithed average over locations' values

    • ss standard deviation of sensitivity calculated as the unweithed standard deviation over locations' values

    • md mean deviation (MD) calculated as the weithed average over locations' values. Weights are the inverse of the standard deviation in sd for TD at each location.

    • sd standard deviation of total deviation calculated as the weithed standard deviation over locations' values. Weights are the inverse of the standard deviation in sd for TD at each location.

    • pmd pattern mean deviation calculated as the weithed average over locations' values. Weights are the inverse of the standard deviation in sd for PD at each location.

    • psd pattern standard deviation calculated as the weithed standard deviation over locations' values. Weights are the inverse of the standard deviation in sd for PD at each location.

    • gh general height. This is defined traditionally for the 24-2 and the 30-2 as the approximatelly the 85th percentile of TD values

    • vfi the oddly defined visual field index

• tdfun a function defining how to obtain the TD values. Typically, it is a function of age and sensitivity values and it is defined as sensitivity values minus the age-corrected mean normal obtained as defined in agem. Thus, TD values are negative is visual field sensitivity values are below mean normal and positive if they are above mean normal

• ghfun a function defining how to obtain the general height

• pdfun a function defining how to obtain the PD values. Tipically, they are obtaines as the TD values minus the general height

• glfun a function defining how to obtain different global indices

• tdpfun, pdpfun, glpfun mapping functions to get the probability levels corresponding to TD, PD and global indices values and based on the lookup tables defined in luts

Examples

# generate normative values from SUNY-IU dataset of healthy eyes
# pointwise
sunyiu_24d2_pw <- nvgenerate(vfctrSunyiu24d2, method = "pointwise",
                             name      = "SUNY-IU pointwise NVs",
                             perimetry = "static automated perimetry",
                             strategy  = "SITA standard",
                             size      = "Size III")
# smooth
sunyiu_24d2 <- nvgenerate(vfctrSunyiu24d2, method = "smooth",
                          name      = "SUNY-IU smoothed NVs",
                          perimetry = "static automated perimetry",
                          strategy  = "SITA standard",
                          size      = "Size III")

visualFields documentation built on March 18, 2022, 7:14 p.m.