reproducer (version 0.5.3)

metaanalyse.Cliffd: metaanalyse.Cliffd

Description

This function provides a simple meta-analysis of experiments using Cliff's d as an effect size. It returns the 100*(1-alpha/2)

Usage

metaanalyse.Cliffd(
  Cliffd,
  Cliffdvar,
  df = 0,
  alternative = "two.sided",
  alpha = 0.05
)

Value

Estimate The overall estimate of Cliff's d obtained from the set of experiments

UpperCI The upper 100*(1-alpha/2)

LowerCI The lower 100*(1-alpha/2)

The variance of the Estimate

tvalue The value of the t-statistic

df The supplied degrees of freedom or NA if the input parameter df was set to zero

AltHyp Defines the alternative hypothesis used for significance testing and depends on the value of the input parameter alternative. It takes the values "Not=0", ">0", or "<0"

NullHyp Defines the null hypothesis and depends on the value of the input parameter alternative. It takes the values "~0", "<0", or ">0"

pvalue The p-value of the t-test if the parameter df>0, or the normal probability value if d=0

RejectNullHyp "Yes" or "No" depending on whether or not the null hypothesis should be rejected at the alpha/2 level for two-sided tests and alpha level for one-sided tests

The Q homogeneity statistic

The I-squared estimate of the extent of heterogeneity

ProbQHomogeneous. The probability that the set of Cliff's d values come from a set of homogeneous experiments.

Arguments

Cliffd

A vector of one or more numerical values, identifying the effect sizes to be meta-analysed

Cliffdvar

A vector of the estimates variance of each of the effect sizes

df

The total degrees of freedom for the set of effect sizes. If df>0, the pvalues and significance test use the t-distribution probability values. If df=0 (default) the pvalues and significance test use the normal distribution probability values. The confidence intervals are always based on the normal probability values, as recommended by Cliff.

alternative

Specifies the type of significance test and can take the values "two.sided" (default), "less" or "greater".

alpha

The significance level used to control the significance tests and calculation of confidence limits (default 0.05).

Author

Barbara Kitchenham and Lech Madeyski

Examples

Cliffd=c(0.84,0.2,-0.04,0.44,0.76)
CliffdvarInvalid=c(0.04,0.18,0.21,0.15)
Cliffdvar=c(0.04,0.18,0.21,0.15,0.06)
CliffdvarInvalid=c(0.04,0.18,0.21,0.15)
df=45
as.data.frame(metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=Cliffdvar,df=df,alternative="greater",
 alpha=0.05))
#  Estimate   UpperCI   LowerCI Variance tvalue df AltHyp NullHyp      pvalue
#1     0.44 0.6601381 0.1502568   0.0256   2.75 45     >0     <=0 0.004275955
#  RejectNullHyp    Q I.square ProbQHomogeneous
#1           Yes 21.5 81.39535     0.0002519835
as.data.frame(metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=Cliffdvar,df=df,alternative="less",
 alpha=0.05))
#  Estimate   UpperCI   LowerCI Variance tvalue df AltHyp NullHyp   pvalue RejectNullHyp
#1     0.44 0.6601381 0.1502568   0.0256   2.75 45     <0     >=0 0.995724            No
#     Q I.square ProbQHomogeneous
#1 21.5 81.39535     0.0002519835
as.data.frame(metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=Cliffdvar,df=df,alternative="two.sided",
 alpha=0.05))
#  Estimate  UpperCI    LowerCI Variance tvalue df AltHyp NullHyp      pvalue
#1     0.44 0.692073 0.09227496   0.0256   2.75 45  Not=0      ~0 0.008551911
#  RejectNullHyp    Q I.square ProbQHomogeneous
#1           Yes 21.5 81.39535     0.0002519835
as.data.frame(metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=Cliffdvar,df=df,alpha=0.05))
#  Estimate  UpperCI    LowerCI Variance tvalue df AltHyp NullHyp      pvalue
#1     0.44 0.692073 0.09227496   0.0256   2.75 45  Not=0      ~0 0.008551911
#  RejectNullHyp    Q I.square ProbQHomogeneous
#1           Yes 21.5 81.39535     0.0002519835
metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=Cliffdvar,df=0,alternative="two.sided",alpha=0.05)
#Error in testfunctionParameterChecks(alternative = alternative, alpha = alpha,  :
#  Invalid alternative parameter, choose one of two.sided, greater or less
# metaanalyse.Cliffd(Cliffd=Cliffd,Cliffdvar=CliffdvarInvalid,df=df,alternative="greater",
# alpha=0.05)
#Error in metaanalyse.Cliffd(Cliffd = Cliffd, Cliffdvar = CliffdvarInvalid,  :
#  Length of Cliffdvar parameter must equal the length of the Cliffd parameter