Skip to content

Commit

Permalink
Included Chisquare-test and ANOVA tabs for power calculation
Browse files Browse the repository at this point in the history
  • Loading branch information
Nils Kappelmann committed Jan 2, 2020
1 parent 68ab001 commit beb1c6e
Show file tree
Hide file tree
Showing 2 changed files with 482 additions and 46 deletions.
242 changes: 215 additions & 27 deletions server.R
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,86 +12,166 @@ library("pwr")
server <- function(input, output) {


# ---------------General npRCT function-------------
# ---------------General npRCT functions-------------

## T-test

npRCT <- reactive({
npRCT.t <- reactive({

# Run power calculation for first t-test (traditional part of RCST)
t1 <- pwr.t.test(d = input$d1,
sig.level = input$sig.level1,
power = input$power1,
alternative = input$alternative1,
t1 <- pwr.t.test(d = input$t.d1,
sig.level = input$t.sig.level1,
power = input$t.power1,
alternative = input$t.alternative1,
type = "two.sample")

# Run power calculation for second t-test (stratified part of RCST)
t2 <- pwr.t.test(d = input$d2, sig.level = input$sig.level2, power = input$power2,
t2 <- pwr.t.test(d = input$t.d2, sig.level = input$t.sig.level2,
power = input$t.power2,
type = "two.sample",
alternative = input$alternative2)
alternative = input$t.alternative2)

# Compute sample sizes (per group)
n1 = ceiling(t1$n)
n2 = t2$n
n2_rct = ceiling(n2 / 2) # n still traditionally randomised in stratified part

# Compute output
n_strat = n1 - n2_rct
n_total = n_strat + n2
n_trad = n1 - n2_rct
n_total = n_trad + n2
n_saved = (n1 + n2) - n_total
n_ind = n1 + n2

# Save output
npRCT <- data.frame(n_total = round(n_total),
n_strat = n_strat,
npRCT.t <- data.frame(n_total = round(n_total),
n_trad = n_trad,
n_saved = n_saved,
n_ind = n_ind)

# Return output
npRCT
npRCT.t

})


## Chisquare test

npRCT.chisquare <- reactive({

# Run power calculation for first chisquare-test (traditional part of npRCT)
chisq1 <- pwr.chisq.test(w = input$chisquare.w1, sig.level = input$chisquare.sig.level1,
power = input$chisquare.power1,
df = (input$chisquare.k-1)*(input$chisquare.ycat-1))

# Run power calculation for second chisquare-test (stratified part of npRCT)
chisq2 <- pwr.chisq.test(w = input$chisquare.w2, sig.level = input$chisquare.sig.level2,
power = input$chisquare.power2, df = (input$chisquare.ycat-1))

# Compute sample sizes (per group)
n1 = ceiling(chisq1$N / input$chisquare.k)
n2 = chisq2$N / 2
n2_rct = ceiling(n2 / input$chisquare.k) # n still traditionally randomised in stratified part

n_trad = n1 - n2_rct
n_total = ceiling(n_trad + n2)
n_ind = n1 + n2
n_saved = n_ind - n_total



# Save output
npRCT.chisquare <- data.frame(n_total = n_total,
n_trad = n_trad,
n_saved = n_saved,
n_ind = n_ind)

# Return output
npRCT.chisquare

})


## ANOVA

npRCT.anova <- reactive({


# Run power calculation for first t-test (traditional part of npRCT)
anova1 <- pwr.anova.test(k = input$anova.k, f = input$anova.f1,
sig.level = input$anova.sig.level1, power = input$anova.power1)

# Run power calculation for second t-test (stratified part of npRCT)
t2 <- pwr.t.test(d = input$anova.d2, sig.level = input$anova.sig.level2,
power = input$anova.power2, type = input$anova.type2,
alternative = input$anova.alternative2)

# Compute sample sizes (per group)
n1 = ceiling(anova1$n)
n2 = t2$n
n2_rct = ceiling(n2 / input$anova.k) # n still traditionally randomised in stratified part

n_trad = n1 - n2_rct
n_total = n_trad + n2
n_ind = n1 + n2
n_saved = n_ind - n_total


# Save output
npRCT.anova <- data.frame(n_total = round(n_total),
n_trad = n_trad,
n_saved = n_saved,
n_ind = n_ind)

# Return output
npRCT.anova

})



# ---------------Define output as text--------------

output$n_total <- renderText({
## T-test

output$t.n_total <- renderText({

temp <- npRCT()
temp <- npRCT.t()
paste("Total npRCT sample size: ",
temp$n_total * 2, sep = "")

})

output$n_traditional <- renderText({
output$t.n_traditional <- renderText({

temp <- npRCT()
temp <- npRCT.t()
paste("Traditional RCT sample size (per group): ",
temp$n_strat, sep = "")
temp$n_trad, sep = "")

})


output$n_precision <- renderText({
output$t.n_precision <- renderText({

temp <- npRCT()
temp <- npRCT.t()
paste("Precision RCT sample size (per group): ",
as.character(temp$n_total - temp$n_strat), sep = "")
as.character(temp$n_total - temp$n_trad), sep = "")

})

output$n_saved <- renderText({
output$t.n_saved <- renderText({

temp <- npRCT()
temp <- npRCT.t()
as.character(temp$n_saved)
paste("Using the npRCT design (as compared to two independent RCTs) thus saves a sample size of ",
temp$n_saved, " participants per group.", sep = "")
})

output$warning <- renderText({
output$twarning <- renderText({

temp <- npRCT()
temp <- npRCT.t()

if(temp$n_strat < 20) {
if(temp$n_strat >= 0) {
if(temp$n_trad < 20) {
if(temp$n_trad >= 0) {
as.character("Please note that the sample size (per group) of the traditional RCT is relatively small (i.e., smaller than 20). If you aim for online identification of a precision algorithm, you may need to adjust your parameters. Visit the Explanations tab for more insights on practical considerations.")
} else {
as.character("WARNING: You have obtained a negative sample size for the traditional RCT. This occurs, if the to-be-detected effect of the precision RCT is much smaller than for the traditional RCT (i.e., all participants required for testing the traditional research question (intervention A or B) are recruited as part of the precision RCT. If you want to use the npRCT, please adjust parameters accordingly and visit the Explanations tab for more insights on practical considerations.")
Expand All @@ -104,4 +184,112 @@ server <- function(input, output) {
})


## Chisquare

output$chisquare.n_total <- renderText({

temp <- npRCT.chisquare()
paste("Total npRCT sample size: ",
temp$n_total * 2, sep = "")

})

output$chisquare.n_traditional <- renderText({

temp <- npRCT.chisquare()
paste("Traditional RCT sample size (per group): ",
temp$n_trad, sep = "")

})


output$chisquare.n_precision <- renderText({

temp <- npRCT.chisquare()
paste("Precision RCT sample size (per group): ",
as.character(temp$n_total - temp$n_trad), sep = "")

})

output$chisquare.n_saved <- renderText({

temp <- npRCT.chisquare()
as.character(temp$n_saved)
paste("Using the npRCT design (as compared to two independent RCTs) thus saves a sample size of ",
temp$n_saved, " participants per group.", sep = "")
})

output$chisquarewarning <- renderText({

temp <- npRCT.chisquare()

if(temp$n_trad < 20) {
if(temp$n_trad >= 0) {
as.character("Please note that the sample size (per group) of the traditional RCT is relatively small (i.e., smaller than 20). If you aim for online identification of a precision algorithm, you may need to adjust your parameters. Visit the Explanations tab for more insights on practical considerations.")
} else {
as.character("WARNING: You have obtained a negative sample size for the traditional RCT. This occurs, if the to-be-detected effect of the precision RCT is much smaller than for the traditional RCT (i.e., all participants required for testing the traditional research question (intervention A or B) are recruited as part of the precision RCT. If you want to use the npRCT, please adjust parameters accordingly and visit the Explanations tab for more insights on practical considerations.")
}
} else {
as.character("")
}


})



## ANOVA

output$anova.n_total <- renderText({

temp <- npRCT.anova()
paste("Total npRCT sample size: ",
temp$n_total * 2, sep = "")

})

output$anova.n_traditional <- renderText({

temp <- npRCT.anova()
paste("Traditional RCT sample size (per group): ",
temp$n_trad, sep = "")

})


output$anova.n_precision <- renderText({

temp <- npRCT.anova()
paste("Precision RCT sample size (per group): ",
as.character(temp$n_total - temp$n_trad), sep = "")

})

output$anova.n_saved <- renderText({

temp <- npRCT.anova()
as.character(temp$n_saved)
paste("Using the npRCT design (as compared to two independent RCTs) thus saves a sample size of ",
temp$n_saved, " participants per group.", sep = "")
})

output$anovawarning <- renderText({

temp <- npRCT.anova()

if(temp$n_trad < 20) {
if(temp$n_trad >= 0) {
as.character("Please note that the sample size (per group) of the traditional RCT is relatively small (i.e., smaller than 20). If you aim for online identification of a precision algorithm, you may need to adjust your parameters. Visit the Explanations tab for more insights on practical considerations.")
} else {
as.character("WARNING: You have obtained a negative sample size for the traditional RCT. This occurs, if the to-be-detected effect of the precision RCT is much smaller than for the traditional RCT (i.e., all participants required for testing the traditional research question (intervention A or B) are recruited as part of the precision RCT. If you want to use the npRCT, please adjust parameters accordingly and visit the Explanations tab for more insights on practical considerations.")
}
} else {
as.character("")
}


})



}
Loading

0 comments on commit beb1c6e

Please sign in to comment.