Skip to content

Commit 0ee3f24

Browse files
NightlordTWNightlordTW
committed
add narrative summary
1 parent dfa7769 commit 0ee3f24

File tree

2 files changed

+276
-129
lines changed

2 files changed

+276
-129
lines changed

‎inst/app/app.R‎

Lines changed: 181 additions & 129 deletions
Original file line numberDiff line numberDiff line change
@@ -25,166 +25,203 @@ source_dir(file.path(app_dir, "R"))
2525
source_dir(file.path(app_dir, "modules"))
2626

2727

28-
# ---- small helper: get Pr(NI) across B trials for plotting ------------------
29-
prNI_draws_conj <- function(B, p_c, p_t, n_c, n_t, M,
30-
prior = c("flat","power"), prior_args = list(),
31-
n_draws = 2000, seed = NULL) {
32-
prior <- match.arg(prior)
33-
if (!is.null(seed)) set.seed(seed)
34-
pr <- numeric(B)
35-
for (b in seq_len(B)) {
36-
res <- bayesNI_trial_betaBinom_conj(
37-
p_c = p_c, p_t = p_t, n_c = n_c, n_t = n_t, M = M,
38-
prior = prior, prior_args = prior_args,
39-
n_draws = n_draws
40-
)
41-
pr[b] <- unname(res$summary["post_prob_NI"])
42-
}
43-
pr
44-
}
45-
46-
4728
# If your functions live in a package, uncomment:
4829
# library(bcts)
4930

50-
ui <- fluidPage(
51-
withMathJax(), # enable LaTeX rendering for the whole page
52-
titlePanel("Bayesian Trial Simulation (Beta–Binomial, conjugate)"),
53-
sidebarLayout(
54-
sidebarPanel(
55-
56-
# --- Assumptions (truth used for simulation / planning) ---
57-
wellPanel(
58-
h4("Treatment Arm Assumptions"),
59-
sliderInput("pt", HTML("True response rate (\\(\\theta_t\\))"),
60-
min = 0, max = 100, value = 85, step = 1,
61-
post = "%"),
62-
numericInput("nt", HTML("Number of randomized patients (\\(n_t \\))"), value = 29, min = 1, step = 1),
31+
ui <- navbarPage(
32+
title = "Bayesian Trial Simulation (Beta–Binomial, conjugate)",
33+
tabPanel(
34+
title = "Randomized trial",
35+
fluidPage(
36+
withMathJax(), # enable LaTeX rendering for the whole page
37+
38+
# --- Description row at top (full width) ---
39+
fluidRow(
40+
column(
41+
width = 12,
42+
div(
43+
style = "margin-bottom: 20px;",
44+
h4("Simulate binary responder outcomes from a randomized trial with a treatment and control arm."),
45+
p("Use the controls in the left panel to define the data-generating assumptions, prior distributions, and decision criteria.
46+
The design summary and operating characteristics will be shown on the right.")
47+
)
48+
)
6349
),
6450

65-
# --- Assumptions (truth used for simulation / planning) ---
66-
tagList(
67-
# Panel 1: Control arm (always shown)
51+
52+
sidebarPanel(
53+
6854
wellPanel(
69-
h4("Control Arm Assumptions"),
70-
sliderInput("pc", HTML("True response rate (\\(\\theta_c\\))"),
71-
min = 0, max = 100, value = 85, step = 1, post = "%"),
72-
numericInput("nc", HTML("Number of randomized patients (\\(n_c\\))"),
73-
value = 29, min = 1, step = 1),
74-
75-
selectInput(
76-
"prior",
77-
"Prior distribution",
78-
choices = c("Flat (no external evidence)" = "flat",
79-
"Power prior (with historical data)" = "power"),
80-
selected = "flat"
55+
h4("Summary Output Type"),
56+
radioButtons(
57+
inputId = "summary_type",
58+
label = "Choose summary format:",
59+
choices = c("Narrative" = "narrative", "Technical" = "technical"),
60+
selected = "narrative"
8161
)
8262
),
8363

84-
# Panel 2: External data config (only when prior == power)
85-
conditionalPanel(
86-
condition = "input.prior == 'power'",
87-
wellPanel(
88-
h4("External Data (Power Prior)"),
89-
sliderInput("a0", "Discount factor a₀ (0 = ignore history, 1 = full borrow)",
90-
min = 0, max = 100, value = 50, step = 1, post = "%"),
91-
numericInput("y0", "Historical responders (y₀)", value = 64, min = 0, step = 1),
92-
numericInput("n0", "Historical sample size (n₀)", value = 75, min = 1, step = 1),
93-
numericInput("abase", "Baseline Beta a_base", value = 1, min = 0.001, step = 0.1),
94-
numericInput("bbase", "Baseline Beta b_base", value = 1, min = 0.001, step = 0.1)
95-
)
96-
)
97-
),
64+
# --- Assumptions (truth used for simulation / planning) ---
65+
wellPanel(
66+
h4("Treatment Arm Assumptions"),
67+
sliderInput("pt", HTML("True response rate (\\(\\theta_t\\))"),
68+
min = 0, max = 100, value = 85, step = 1,
69+
post = "%"),
70+
numericInput("nt", HTML("Number of randomized patients (\\(n_t \\))"), value = 29, min = 1, step = 1),
71+
),
9872

73+
# --- Assumptions (truth used for simulation / planning) ---
74+
tagList(
75+
# Panel 1: Control arm (always shown)
76+
wellPanel(
77+
h4("Control Arm Assumptions"),
78+
sliderInput("pc", HTML("True response rate (\\(\\theta_c\\))"),
79+
min = 0, max = 100, value = 85, step = 1, post = "%"),
80+
numericInput("nc", HTML("Number of randomized patients (\\(n_c\\))"),
81+
value = 29, min = 1, step = 1),
82+
83+
selectInput(
84+
"prior",
85+
"Prior distribution",
86+
choices = c("Flat (no external evidence)" = "flat",
87+
"Power prior (with historical data)" = "power"),
88+
selected = "flat"
89+
)
90+
),
9991

100-
wellPanel(
101-
h4("Decision criteria"),
102-
sliderInput(
103-
"M",
104-
label = HTML("Decision Margin (\\( \\Delta \\))"),
105-
min = -100, max = 100, value = -20, step = 1, post = "%"
106-
),
107-
helpText(
108-
"Δ < 0: non-inferiority (treatment may be up to |Δ| worse).",
109-
"Δ ≥ 0: superiority (treatment must be at least Δ better).",
110-
"Assumes higher response rates are better (responder events)."
92+
# Panel 2: External data config (only when prior == power)
93+
conditionalPanel(
94+
condition = "input.prior == 'power'",
95+
wellPanel(
96+
h4("External Data (Power Prior)"),
97+
sliderInput("a0", "Discount factor a₀ (0 = ignore history, 1 = full borrow)",
98+
min = 0, max = 100, value = 50, step = 1, post = "%"),
99+
numericInput("y0", "Historical responders (y₀)", value = 64, min = 0, step = 1),
100+
numericInput("n0", "Historical sample size (n₀)", value = 75, min = 1, step = 1),
101+
numericInput("abase", "Baseline Beta a_base", value = 1, min = 0.001, step = 0.1),
102+
numericInput("bbase", "Baseline Beta b_base", value = 1, min = 0.001, step = 0.1)
103+
)
104+
)
111105
),
112106

113-
# Choice: set gamma directly OR set alpha and calibrate gamma
114-
radioButtons("decision_mode", "Threshold specification:",
115-
choices = c("Specify posterior probability threshold γ" = "gamma",
116-
"Specify target Type-I error α" = "alpha"),
117-
selected = "gamma"),
118107

119-
conditionalPanel(
120-
condition = "input.decision_mode == 'gamma'",
108+
wellPanel(
109+
h4("Decision criteria"),
121110
sliderInput(
122-
"gamma",
123-
label = HTML("Posterior probability threshold (\\( \\gamma \\))"),
124-
min = 80, max = 99, value = 90, step = 1, post = "%"
111+
"M",
112+
label = HTML("Decision Margin (\\( \\Delta \\))"),
113+
min = -100, max = 100, value = -20, step = 1, post = "%"
125114
),
126-
uiOutput("decision_rule") # placeholder for dynamic help text
127-
),
128-
129-
conditionalPanel(
130-
condition = "input.decision_mode == 'alpha'",
131-
sliderInput(
132-
"alpha",
133-
label = HTML("Target Type-I error (\\( \\alpha \\))"),
134-
value = 10, min = 1, max = 20, step = 1, post = "%"
115+
helpText(
116+
"Δ < 0: non-inferiority (treatment may be up to |Δ| worse).",
117+
"Δ ≥ 0: superiority (treatment must be at least Δ better).",
118+
"Assumes higher response rates are better (responder events)."
135119
),
136-
selectInput(
137-
"calibrate_on",
138-
"Calibrate Type-I on:",
139-
choices = c(
140-
"Point estimate Pr(reject | H₀)" = "point",
141-
"Upper 95% MC CI (conservative)" = "upper",
142-
"Lower 95% MC CI (liberal)" = "lower"
120+
121+
# Choice: set gamma directly OR set alpha and calibrate gamma
122+
radioButtons("decision_mode", "Threshold specification:",
123+
choices = c("Specify posterior probability threshold γ" = "gamma",
124+
"Specify target Type-I error α" = "alpha"),
125+
selected = "gamma"),
126+
127+
conditionalPanel(
128+
condition = "input.decision_mode == 'gamma'",
129+
sliderInput(
130+
"gamma",
131+
label = HTML("Posterior probability threshold (\\( \\gamma \\))"),
132+
min = 80, max = 99, value = 90, step = 1, post = "%"
143133
),
144-
selected = "upper"
134+
uiOutput("decision_rule") # placeholder for dynamic help text
145135
),
146-
helpText("γ will be calibrated so that the chosen Type-I metric ≈ α (within tolerance) at the least-favourable null.")
147-
)
148-
),
136+
137+
conditionalPanel(
138+
condition = "input.decision_mode == 'alpha'",
139+
sliderInput(
140+
"alpha",
141+
label = HTML("Target Type-I error (\\( \\alpha \\))"),
142+
value = 10, min = 1, max = 20, step = 1, post = "%"
143+
),
144+
selectInput(
145+
"calibrate_on",
146+
"Calibrate Type-I on:",
147+
choices = c(
148+
"Point estimate Pr(reject | H₀)" = "point",
149+
"Upper 95% MC CI (conservative)" = "upper",
150+
"Lower 95% MC CI (liberal)" = "lower"
151+
),
152+
selected = "upper"
153+
),
154+
helpText("γ will be calibrated so that the chosen Type-I metric ≈ α (within tolerance) at the least-favourable null.")
155+
)
156+
),
149157

150158

151159

152-
# --- Posterior Evaluation ---
153-
wellPanel(
154-
h4("Simulation settings"),
155-
numericInput("B",
156-
"Number of simulated trials (for Type-I & Power)",
157-
value = 2500, min = 100, step = 100),
158-
helpText(textOutput("design_mc_precision_text")),
160+
# --- Posterior Evaluation ---
161+
wellPanel(
162+
h4("Simulation settings"),
163+
numericInput("B",
164+
"Number of simulated trials (for Type-I & Power)",
165+
value = 2500, min = 100, step = 100),
166+
helpText(textOutput("design_mc_precision_text")),
159167

160-
numericInput("ndraws", "Posterior draws per trial (for Pr(NI))", value = 2000, min = 200, step = 100),
161-
helpText(textOutput("post_mc_precision_text")),
168+
numericInput("ndraws", "Posterior draws per trial (for Pr(NI))", value = 2000, min = 200, step = 100),
169+
helpText(textOutput("post_mc_precision_text")),
162170

163-
numericInput("seed", "Seed (optional)", value = 123, min = 1, step = 1),
164-
#helpText(HTML("Larger values give more precise results but increase runtime.")),
165-
),
171+
numericInput("seed", "Seed (optional)", value = 123, min = 1, step = 1),
172+
#helpText(HTML("Larger values give more precise results but increase runtime.")),
173+
),
166174

167175

168176

169-
actionButton("run", "Run simulation", class = "btn-primary"),
177+
actionButton("run", "Run simulation", class = "btn-primary"),
170178

171-
hr(),
172-
tags$small(
173-
paste("bcts version:", utils::packageVersion("bcts"))
174-
)
175-
),
176-
mainPanel(
177-
mod_designsummary_ui("dsum"),
178-
mod_armpriors_ui("armpriors", height = 320),
179-
mod_oc_ui("oc"),
179+
hr(),
180+
tags$small(
181+
paste("bcts version:", utils::packageVersion("bcts"))
182+
)
183+
),
184+
mainPanel(
185+
conditionalPanel(
186+
condition = "input.summary_type == 'technical'",
187+
mod_designsummary_ui("dsum")
188+
),
189+
conditionalPanel(
190+
condition = "input.summary_type == 'narrative'",
191+
mod_narrative_ui("narrative")
192+
),
193+
mod_armpriors_ui("armpriors", height = 320),
194+
mod_oc_ui("oc"),
180195

181-
# --- Sensitivity analysis ---
182-
# Sidebar
183-
mod_sensitivity_sidebar_ui("sens"),
196+
# --- Sensitivity analysis ---
197+
# Sidebar
198+
mod_sensitivity_sidebar_ui("sens"),
184199

185-
# Main panel (where you want outputs)
186-
mod_sensitivity_main_ui("sens", plot_height = 300)
200+
# Main panel (where you want outputs)
201+
mod_sensitivity_main_ui("sens", plot_height = 300)
187202

203+
204+
)
205+
)
206+
),
207+
208+
tabPanel(
209+
title = "Single-arm trial",
210+
fluidPage(
211+
h4("Single-arm trial dashboard (under construction)"),
212+
p("This section will allow evaluation of single-arm designs using Beta–Binomial conjugate models."),
213+
p("You can simulate a posterior for a single group, compare against a threshold, or incorporate external data via power priors."),
214+
br(),
215+
wellPanel(
216+
p("Include UI elements here for:"),
217+
tags$ul(
218+
tags$li("True response rate (θ)"),
219+
tags$li("Sample size (n)"),
220+
tags$li("Prior type: Flat vs Power prior"),
221+
tags$li("External data if using power prior"),
222+
tags$li("Decision margin (Δ) and threshold (γ or α)")
223+
)
224+
)
188225
)
189226
)
190227
)
@@ -279,6 +316,21 @@ server <- function(input, output, session) {
279316
decision_mode = reactive(input$decision_mode)
280317
)
281318

319+
mod_narrative_server(
320+
"narrative",
321+
pt = reactive(input$pt / 100), nt = reactive(input$nt),
322+
pc = reactive(input$pc / 100), nc = reactive(input$nc),
323+
M = reactive(input$M / 100),
324+
prior = reactive(input$prior),
325+
prior_args = prior_args,
326+
decision_mode = reactive(input$decision_mode),
327+
gamma = reactive(input$gamma / 100),
328+
alpha = reactive(input$alpha / 100),
329+
B = reactive(input$B),
330+
ndraws = reactive(input$ndraws),
331+
seed = reactive(input$seed)
332+
)
333+
282334

283335
mod_designsummary_server(
284336
"dsum",

0 commit comments

Comments
 (0)