DerSimonian and Laird-based Random effect model for standard meta-analysis of risk estimate (e.g relative risk (RR), odds ratio (OR) or hazard ratio (HR))
priskran(rr, u, l, form = c("Log", "nonLog"), conf.level = 0.95)
| rr | A numeric vector of the risk estimated from the individual studies |
|---|---|
| u | A numeric vector of the upper bound of the confidence interval of the risk reported from the individual studies. |
| l | A numeric vector of the lower bound of the confidence interval of the risk reported from the individual studies. |
| form | Logical, indicating the scale of the data. If Log, then the original data are in logarithme scale. |
| conf.level | Coverage for confidence interval |
Object of class "metaan.ra". A list that print the output from the priskran function. The following could be found from the list :
rr_tot (Effect): The pooled effect from the individual studies' estimate (RR, OR, or HR)
sd_tot_lnRR (SE-Log(Effect)): The standard error of the pooled effect (see reference Richardson et al 2020 for more details)
l_tot (Lower CI): The lower confidence interval bound of the pooled effect (rr_tot)
u_tot (Upper CI): The upper confidence interval bound of the pooled effect (rr_tot)
Cochrane_stat (Cochran’s Q statistic): The value of the Cochrane's statistic of inter-study heterogeneity
Degree_freedom (Degree of Freedom): The degree of freedom
p_value (P-Value): The p-value of the statistic of Cochrane
I_square (Higgins’ and Thompson’s I^2 (%)): I square value in percent (%) indicating the amount of the inter-study heterogeneity
DerSimonian R, Laird N (1986) Meta-analysis in clinical trials. Controlled clinical trials 7:177–188.
study <- c("Canada", "Northern USA", "Chicago", "Georgia","Puerto", "Comm", "Madanapalle", "UK", "South Africa", "Haiti", "Madras") Risk <- c(0.205, 0.411, 0.254, 1.562, 0.712, 0.983, 0.804, 0.237, 0.625, 0.198, 1.012) lower_ci <- c(0.086, 0.134, 0.149, 0.374, 0.573, 0.582, 0.516, 0.179, 0.393, 0.078, 0.895) upper_ci <- c(0.486, 1.257, 0.431, 6.528, 0.886, 1.659, 1.254, 0.312, 0.996, 0.499, 1.145) donne <- data.frame(cbind(study, Risk, lower_ci, upper_ci)) donne$Risk <- as.numeric(as.character(donne$Risk)) donne$upper_ci <- as.numeric(as.character(donne$upper_ci)) donne$lower_ci <- as.numeric(as.character(donne$lower_ci)) # on the log form donne$ln_risk <- log(donne$Risk) donne$ln_lower_ci <- log(donne$lower_ci) donne$ln_upper_ci <- log(donne$upper_ci) priskran(rr=donne$Risk, u=donne$upper_ci, l=donne$lower_ci, form="nonLog", conf.level=0.95)#> #> Standard meta-analysis with random effect model #> -------------------------------------------------- #> #> Effect SE-Log(Effect) Lower CI Upper CI #> 0.51 0.21 0.34 0.77 #> #> -------------------------- ----------------------- #> #> Test of heterogeneity : #> #> Cochran Q statistic Degree of Freedom P-Value #> 125.05 10.00 0 #> #> -------------------------------------------------- #> #> Higgins and Thompson I^2 (%) #> 92 #> __________________________________________________ #>priskran(rr=donne$ln_risk, u=donne$ln_upper_ci, l=donne$ln_lower_ci, form="Log", conf.level=0.95)#> #> Standard meta-analysis with random effect model #> -------------------------------------------------- #> #> Effect SE-Log(Effect) Lower CI Upper CI #> 0.51 0.21 0.34 0.77 #> #> -------------------------- ----------------------- #> #> Test of heterogeneity : #> #> Cochran Q statistic Degree of Freedom P-Value #> 125.05 10.00 0 #> #> -------------------------------------------------- #> #> Higgins and Thompson I^2 (%) #> 92 #> __________________________________________________ #>