clinsig (version 1.2)

clinsig: Calculate clinical significance criteria.

Description

Calculates one or more parametric or non-parametric criteria for clinical significance.

Usage

clinsig(pre.x,post.x,dys.mct=NA,func.mct=NA,dys.disp=NA,func.disp=NA, mct="mean",disp="sd",disp.mult=c(2,2),dys.qnts=NA, dir.effect=NA,xlim=range(c(pre.x,post.x),na.rm=TRUE), ylim=range(c(pre.x,post.x),na.rm=TRUE),pch=par("pch"), do.plot=TRUE,point.id=NA,which.crit="c",coef.alpha=NA,rc.mult=1.96,...)

Arguments

pre.x

A vector of pre-intervention assessment scores.

post.x

A vector of post-intervention assessment scores for the same measure as pre.x.

dys.mct

A Measure of Central Tendency (MCT) for the scores of the population of dysfunctional persons on the measure used. If absent, it will be estimated from pre.x.

func.mct

A Measure of Central Tendency for the normative ("functional") scores on the measure.

dys.disp

A measure of DISPersion for the scores of the population of dysfunctional persons on the measure. If missing, it will also be estimated.

func.disp

A measure of DISPersion for the normative scores on the measure.

mct

The name of a Measure of Central Tendency to be used. Defaults to the mean.

disp

The name of a measure of DISPersion to be used. Defaults to the standard deviation.

disp.mult

The multiple(s) of the dispersion measure to be used in calculations. If different multiples are to be used for the pre/post measures and the normative scores, pass a two element vector containing the multipliers for the pre/post and then the normative dispersions. Defaults to 2 for both.

dys.qnts

The lower and upper quantiles of the pre-assessment intervention scores to be used in calculating the "a" criterion. If a two element vector of quantiles such as c(0.16,0.84) is passed, this will be used instead of dys.disp or the dispersion calculated for pre.x. See Details.

dir.effect

The direction of effect expected of the intervention. This argument should only take the values of -1 or 1. see Details.

xlim,ylim

Allows the user to set the x and y axes to the range desired. Defaults to range(pre.x) and range(post.x) respectively.

pch

Symbols to use in the plot.

do.plot

Whether to display a significance plot.

point.id

Optional text to be displayed instead of symbols for each score pair.

which.crit

Which criterion should be used if reliable change measure is to be calculated.

coef.alpha

sample Cronbach's alpha for the instrument used to measure the dysfunction. If NOT NA, this causes the reliable change measure to be calculated.

rc.mult

The multiplier for the standard error in the reliable change measure. See Details.

...

Additional arguments passed to plot.

Value

A list containing:As a side effect, a scatterplot of the pre- and post-scores and MCTs with the calculated cutoff scores as lines is displayed if do.plot is TRUE.

Details

The clinsig function calculates both parametric and non-parametric versions of the Jacobson-Truax estimates of clinical significance. The number of estimates produced will depend upon which arguments are supplied. The "a" estimate is always calculated as it only requires the measure of central tendency (MCT) and dispersion of the pre-intervention scores. The c estimate requires that the normal (non-pathologic) measure of central tendency be specified. The "b" estimate requires the dispersion of the normal MCT. If the dispersion of the dysfunctional scores is not supplied, the function will estimate this with the function named in disp. The user can calculate a measure of dispersion and pass it as dys.disp or write a function and pass the name of that function as disp. Note that the use of non- parametric estimates is not included in the work Jacobson and colleagues. Using the "a" cutoff means that post-intervention scores are significant if they are at least a specified distance from the MCT of pre-intervention scores in the direction of the normal MCT (i.e. they are "far enough" from the dysfunctional MCT). The "b" cutoff means that post-intervention scores are significant if they are within a specified distance of the normal MCT (i.e. they are "close enough" to the normal MCT). This criterion is typically used with overlapping distributions. The "c" cutoff means that post-intervention scores are in the direction of the normal MCT and beyond the weighted mean of the two MCTs (i.e. they are "on the normal side" of the average of the MCTs). The "c" criterion is usually recommended for non-overlapping distributions of pre- and post-scores. Look at the second plot in the examples for an illustration of this in which there is a large effect and a clear separation of pre- and post-scores.

The direction of effect is calculated as the sign of the functional (normal) MCT minus the dysfunctional MCT. If one or both of these are missing, the pre- and post-MCTs are substituted. This assumes that the post-MCT is in the expected direction toward the functional MCT. If this is not the case, the user should set dir.effect to 1 (higher scores = improvement) or -1 (lower scores = improvement) or the function may report improvement when the scores are actually indicating deterioration.

If only the pre-intervention and post-intervention scores are provided, the function calculates a cutoff score for significant change based upon disp.mult times the measure of dispersion for the pre-intervention scores if dys.disp is specified. However, if dys.qnts is not NA, the "a" cutoff score is set to one of the quantiles of the pre-intervention score distribution. If the normal MCT is supplied and it is less than the pre-intervention MCT, or dir.effect is -1, the lower quantile is used. In all other cases, the upper cutoff is calculated. Remember that the quantile is not adjusted for disp.mult. The reliable change measure is calculated following Evans, Margison & Barkham (1998). This calculation assumes that the sample score distribution is normal. It will be calculated with the measures specified in sampmct and sampdisp, but unless these are the defaults, it is not guaranteed to be correct.

References

Evans, C., Margison, F. & Barkham, M. (1998) The contribution of reliable and clinically significant change methods to evidence-based mental health. Evidence Based Mental Health, 1: 70-72.

Jacobson, N.S. & Truax, P. (1991) Clinical significance: a statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59(1): 12-19.

Examples

 pre.x<-runif(30,3,6)
 post.x<-runif(30,1,4)
 clinsig(pre.x,post.x,func.mct=1,func.disp=1,xlim=c(1,6),ylim=c(1,6))
 # simulate scores on a typical psychological assessment with a limited
 # range and a large separation between the pre- and post- assessments
 pre.x<-c(3,3,4,5,5,6,6,6,6,7,7,7,8,8,8,8,9,9,10,10)
 post.x<-c(13,12,15,14,12,18,13,17,NA,20,16,22,23,15,19,17,18,21,13,15)
 big.sep<-clinsig(pre.x,post.x,mct="median",disp="mad",func.mct=19,func.disp=2,
  do.plot=FALSE)
 hist(big.sep,main="Widely separated samples")
 legend(20,3.8,c("Pre","Post"),fill=c("red","green"))
 # now squeeze the two samples together to show how the criteria change positions
 post.x<-post.x-7
 little.sep<-clinsig(pre.x,post.x,mct="median",disp="mad",func.mct=15,func.disp=2,
  do.plot=FALSE)
 hist(little.sep,main="Closely spaced samples")
 legend(12.5,3.8,c("Pre","Post"),fill=c("red","green"))
 # example from Evans, Margison & Barkham, 1998 with simulated data
 set.seed(12345)
 # values from EMB 
 pre_mct<-1.81
 pre_disp<-0.53
 post_mct<-0.79
 post_disp<-0.5
 func_mct<-0.72
 func_disp<-0.57
 # accept EMB's normality of distribution
 pre_treat<-rnorm(40,pre_mct,pre_disp)
 post_treat<-rnorm(40,post_mct,post_disp)
 # make sure that no scores go below zero
 post_treat[post_treat<0]<-0
 emb<-clinsig(pre_treat,post_treat,
  dys.mct=pre_mct,func.mct=func_mct,
  dys.disp=pre_disp,func.disp=func_disp,
  coef.alpha=0.89,
  main="Clinical significance plot with reliable change window")
 print(emb)
 hist(emb)