Cohort characteristics

library(mbtools)
theme_set(theme_minimal())

cohort <- rbind(
    fread("no_weight_loss.csv", colClasses=c(public_client_id="character")),
    fread("successful_weight_loss.csv", colClasses=c(public_client_id="character"))
)
cohort[, "subset" := "controls"]
cohort[weight_change_relative < 0, "subset" := "weight loss"]
cohort[, subset := factor(subset)]

subcohort <- fread("data/manifest.csv", colClasses = classes)[has_close_microbiome == T]
subcohort[, "public_client_id" := paste0("0", public_client_id)]

cohort[, "additional_assays" := FALSE]
cohort[public_client_id %chin% subcohort$public_client_id, "additional_assays" := TRUE]

library(arivale.data.interface)
chem <- get_snapshot("chemistries", clean=T)
chem <- chem[cohort, on=c("public_client_id", "days_in_program"), roll="nearest"]

tt <- cohort[, t.test(bmi[since_baseline > 0], bmi[since_baseline == 0], paired=TRUE)]

ggplot(cohort, aes(x=since_baseline, y=bmi, group=public_client_id, 
                   fill=additional_assays, color=subset)) +
    geom_line(alpha=0.5) +
    geom_point(shape=21, stroke=0, size=2.5) +
    scale_fill_manual(values=c("black", "mediumseagreen")) +
    labs(x="time in program [days]", y="BMI [kg/m²]") +
    guides(color="none", fill="none")

ggsave("figures/bmi_trajectories.png", width=4, height=3, dpi=300)

cohort[, t.test(bmi[since_baseline == 0], bmi[since_baseline > 0], paired=TRUE)]

    Paired t-test

data:  bmi[since_baseline == 0] and bmi[since_baseline > 0]
t = 5.2091, df = 104, p-value = 9.647e-07
alternative hypothesis: true difference in means is not equal to 0
95 percent confidence interval:
 0.655356 1.461042
sample estimates:
mean of the differences 
               1.058199 

measures <- c("ADIPONECTIN__SERUM", "HDL_CHOL_DIRECT", "LDL_CHOL_CALCULATION", 
              "GLUCOSE", "HOMA_IR", "INSULIN", "CRP_HIGH_SENSITIVITY", 
              "GLYCOHEMOGLOBIN_A1C")
legends <- c("serum adiponectin [mg/L]", "serum HDL [mg/L]", "serum LDL [mg/L]", 
             "serum glucose [mg/L]", "HOMA IR [index]", "serum insulin [mIU/L]", 
             "serum CRP [mg/L]", "glycated hemoglobin [%]")
select <- chem[, c("public_client_id", "since_baseline", "additional_assays", measures), with=F]
names(select) <- c("public_client_id", "since_baseline", "additional_assays", legends)
select <- melt(select, id.vars=c("public_client_id", "since_baseline", "additional_assays"), value.name="value", variable.name="measure")

stats <- function(value, since_baseline) {
    s <- t.test(value[since_baseline > 0], value[since_baseline == 0], paired=TRUE)
    list(t=s$statistic, p=s$p.value, delta_mu=s$estimate)
}

res <- select[, stats(value, since_baseline), by="measure"]
res[, "label" := sprintf("Δ=%.2g, p=%.2g", delta_mu, p)]

pl <- ggplot(select, aes(x=since_baseline, y=value, group=public_client_id, color=additional_assays)) +
    geom_line(alpha=0.5, color="seagreen4") +
    geom_point() +
    labs(x="time in program [days]", y="") +
    facet_wrap(~ measure, scales="free_y") +
    guides(color="none") +
    scale_color_manual(values=c("black", "tomato"))
pl <- pl + geom_text(data = res, 
                     mapping = aes(x = Inf, y = Inf, label = label, group=1),
                     color = "black",
                     hjust = 1,
                     vjust = 2.5)
ggsave("figures/other_trajectories.png", width=8, height=8, dpi=300)
pl