Code / comments

This template generates output for numerical/continuous outcome.

The following are computed

Summary measures

For each combination of assessment time and randomization group we compute

  • number of observations (excluding missing data!)
  • mean
  • standard deviation
  • median
  • interquantile range (given by 25 % and 75 % quantiles)

For each randomization group summary measures are also computed for changes over time (baseline to follow up).

Model based estimates

A linear mixed model with fixed effects of time, randomization group and their interaction as well as random effect of subject/ID_NUM is used to extract

  • mean estimate for alle combinations of group and time
  • within group changes over time
  • comparison of within group changes over time across the two groups

The tables produced extract

  • estimates
  • standard error on estimates
  • P-value (for testing if estimate equals to zero)
  • lower and upper 95 % confidence interval for estimate

Source Code

sum_na <- function(x){sum(!is.na(x))}
my_funs <- funs(n = sum_na, mean = round(mean(., na.rm = T), digits = 4), sd = round(sd(., na.rm = T), digits = 4), median = median(., na.rm = T), q25 = quantile(., probs = 0.25, na.rm = T), q75 = quantile(., probs = 0.75, na.rm = T))

print_num <- function(varname, df = data_full){
    group_by(df, time, Randomisering) %>% 
    select(time, Randomisering, varname) %>%
    summarize_at(.vars = varname, .funs = my_funs) %>%
    filter(!is.na(Randomisering))
}

Results for variable N2

Here given for variable: N2

Raw data

DT::datatable(print_num(params$var_name))

Model based

library(nlme)
form1 <- formula(paste(params$var_name, "~ time * Randomisering"))
mod1 <- lme(form1, random = ~1| ID_NUM, data = data_full, na.action = na.omit)

form2 <- formula(paste(params$var_name, "~ Randomisering + time:Randomisering - 1"))
mod2 <- lme(form2, random = ~1| ID_NUM, data = data_full, na.action = na.omit)

form3 <- formula(paste(params$var_name, "~ time:Randomisering - 1"))
mod3 <- lme(form3, random = ~1| ID_NUM, data = data_full, na.action = na.omit)

tmp_table3 <- round(cbind(summary(mod3)$tTable[1:4, c(1, 2, 5)], intervals(mod3)$fixed[, c(1, 3)]
), digits = 4)[, c(1, 2, 4, 5, 3)]

tmp_table2 <- round(cbind(summary(mod2)$tTable[3:4, c(1, 2, 5)], intervals(mod2)$fixed[3:4, c(1, 3)]
), digits = 4)[, c(1, 2, 4, 5, 3)]

tmp_table1 <- round(c(summary(mod1)$tTable[4, c(1, 2, 5)], intervals(mod1)$fixed[4, c(1, 3)]
), digits = 4)[c(1, 2, 4, 5, 3)]

tmp_tab <- as_tibble(rbind(tmp_table3, tmp_table2, tmp_table1))
tmp_tab <- mutate(tmp_tab, contrast = c("control_base", "control_follow", "intervention_base", "intervention_follow", "control_change", "intervention_change", "group_diff_change"))[c(6, 1:5)]

DT::datatable(tmp_tab)