calculating medians per year per ID in R and plotting the outcome - r

Question

Dataset:
structure(list(ID = c(1234, 1234, 1234, 1234, 1234, 1234, 1234,
1234, 8769, 8769, 8769, 8769, 8769, 7457, 7457, 7457, 7457, 7457,
7457, 55667, 55667, 55667, 55667, 55667, 55667, 55667, 3789,
3789, 3789, 3789, 3789, 3789), date_of_bloods = structure(c(978307200,
981072000, 1173052800, 1175731200, 1367798400, 1465171200, 1467936000,
1659916800, 1072915200, 1075680000, 1173052800, 1175731200, 1367798400,
978307200, 981072000, 1173052800, 1175731200, 1367798400, 1465171200,
978307200, 981072000, 1173052800, 1270425600, 1273104000, 1465171200,
1467936000, 1270425600, 1367798400, 1465171200, 1465257600, 1465344000,
1465430400), class = c("POSIXct", "POSIXt"), tzone = "UTC"),
result = c(90, 80, 60, 40, 25, 22, 22, 21, 70, 65, 43, 23,
22, 90, 90, 88, 86, 76, 74, 58, 46, 35, 34, 33, 30, 24, 76,
67, 56, 34, 33, 23), `mutation type` = c(1, 1, 1, 1, 1, 1,
1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3,
3, 1, 1, 1, 1, 1, 1)), row.names = c(NA, -32L), class = "data.frame")
I would like the median of results per year per ID in a format where the year is just 0,1,2,3 etc for uniformity across cohorts and then to plot these lines with some indication of their mutation category.
I have done:
filtered$date_of_bloods <-format(filtered$date_of_bloods,format="%Y")
#split into individual ID groups
a <- with(filtered, split(filtered, list(ID)))
#aggregate median results per year
medianfunc <- function(y) {aggregate(results ~ date_of_bloods, data = y, median)}
medians <- sapply(a, medianfunc)
# do lm per ID cohort and get slope of lines
g<- as.data.frame(medians)
coefLM <- function(x) {coef(lm(date_of_bloods ~ results, data = x))}
coefs<- sapply(g, coefLM)
The actual years don't matter and for uniformity I would like them to be 0,1,2,3,4 etc per ID. I am not sure how to do that? I would then want to plot this data (median yearly bloods per ID) with some form of idea as to which mutational category they belong.
I hope this isn't too broad a question.
Many thanks

Answer 1

You can try this (filtered is the dput() you included). I hope this helps:
library(dplyr)
library(lubridate)
library(ggplot2)
library(broom)
#Data
filtered %>% mutate(year=year(date_of_bloods)) %>%
group_by(ID,year,`mutation type`) %>% summarise(med=median(result)) -> df1
#Variables
df1 %>% ungroup()%>% mutate(ID=as.factor(ID),
year=as.factor(year),
`mutation type`=as.factor(`mutation type`)) -> df1
#Plot
ggplot(df1,aes(x=ID,y=med,fill=`mutation type`,color=year,group=year))+
geom_line()
And for models:
#Models
fits <- df1 %>%group_by(ID) %>%
do(fitmodel = lm(med ~ year, data = .))
#Coefs
dfCoef = tidy(fits, fitmodel)
# A tibble: 10 x 6
# Groups: ID [5]
ID term estimate std.error statistic p.value
<dbl> <chr> <dbl> <dbl> <dbl> <dbl>
1 1234 (Intercept) 6329. 1546. 4.09 0.0264
2 1234 year -3.13 0.769 -4.07 0.0268
3 3789 (Intercept) 14318. 4746. 3.02 0.204
4 3789 year -7.08 2.36 -3.00 0.205
5 7457 (Intercept) 2409. 403. 5.98 0.0269
6 7457 year -1.16 0.201 -5.78 0.0287
7 8769 (Intercept) 9268. 4803. 1.93 0.304
8 8769 year -4.60 2.39 -1.92 0.306
9 55667 (Intercept) 3294. 759. 4.34 0.0492
10 55667 year -1.62 0.378 -4.29 0.0503
Code for required plot:
#Plot 2
#Data modifications
df1 %>% mutate(year2=as.numeric(year)-1) -> df2
df2 %>% mutate(year2=factor(year2,levels = sort(unique(year2)))) -> df2
#Plot 2
ggplot(df2,aes(x=year2,y=med,color=ID,group=ID))+
facet_wrap(.~`mutation type`)+
geom_line()

Answer 2

Your naming structure is unclear, if the data you provided is called df then you can do:
df$year <-format(df$date_of_bloods,format="%Y")
aggregate(result ~ year + ID, data = df, median)
year ID result
1 2001 1234 85.0
2 2007 1234 50.0
3 2013 1234 25.0
4 2016 1234 22.0
5 2022 1234 21.0
6 2010 3789 76.0
7 2013 3789 67.0
8 2016 3789 33.5
9 2001 7457 90.0
10 2007 7457 87.0
11 2013 7457 76.0
12 2016 7457 74.0
13 2004 8769 67.5
14 2007 8769 33.0
15 2013 8769 22.0
16 2001 55667 52.0
17 2007 55667 35.0
18 2010 55667 33.5
19 2016 55667 27.0