model.how.new.users.contribute.to.dau.Rmd

```{r}
## Get the JSON file of releases
library(rjson)
library(data.table)
library(infuser)
download.file('https://product-details.mozilla.org/1.0/firefox_history_major_releases.json',dest='/tmp/x.json')
releases <- local({
    x <- fromJSON(file='/tmp/x.json')
    data.table(version=names(x), date=as.Date(unlist(x)))
})[date>='2016-07-01',]


dau4everQ <- sql(" select
 from_unixtime(unix_timestamp(submission_date_s3, 'yyyyMMdd'), 'yyyy-MM-dd')  as date,
 count(distinct(client_id)) as dau
 from main_summary
 where app_name='Firefox' and normalized_channel = 'release' and sample_id='42'
 and submission_date_s3>='20160701'
 group by 1 order by 1")
dau4ever <- data.table(collect(dau4everQ))
dau4ever$date <- as.Date(dau4ever$date)

collector <- list()
for(i in 1:nrow(releases)){
    print(releases[i,])
    q1 <- sql(as.character(infuse(
        "select
       from_unixtime(unix_timestamp(submission_date_s3, 'yyyyMMdd'), 'yyyy-MM-dd')  as date,
       count(distinct(client_id)) as np
     from main_summary
     where app_name = 'Firefox' and normalized_channel = 'release' and sample_id='42'
     and profile_creation_date >= {{dateSinceEpoch}}
     and submission_date_s3  >='{{releasedateYYMMDD}}'
     group by 1 order by 1
    ",dateSinceEpoch=as.integer(releases[i,date]),releasedateYYMMDD=strftime(releases[i,date],"%Y%m%d")
    )))
    x <- data.table(collect(q1))
    x$version <- releases[i,version]
    x$releasedate <- as.Date(releases[i,date])
    x$date <- as.Date(x$date)
    x$daysSince <- as.numeric(x$date - x$releasedate)
    x <- x[daysSince<365,]
    x <- merge(x,dau4ever,by='date',all.x=TRUE)
    x[, pNewOnDAU :=np/dau]
    collector[[ length(collector)+1 ]] <- x
}

y <- rbindlist(collector)
xyplot(pNewOnDAU~ daysSince,type=c('g','l'),col='#00000030',groups=version,data=y)
z=y[, list( mean=mean(pNewOnDAU), low=mean(pNewOnDAU)-1.96*sd(pNewOnDAU)/sqrt(.N), high=mean(pNewOnDAU)+1.96*sd(pNewOnDAU)/sqrt(.N)),                  by=daysSince]