iterate dependent variable using purr and map (tidyverse)

Question

I have a simple dataset that I want to iterate the dependent variable using aov and tidyverse. From those outputs I then want to compute Tukey HSD tests. I have this working in a for loop structure, but am trying my hardest to migrate from that mentality. I saw this post on iterating aovfunctions with the independent variables. Tried to incorporate this logic into my workflow, but not working out so well. Any tidyverse aficionados that could steer me in the right direction here?

library(tidyverse)
library(data.table)

pfuel <- fread("data/CFL.csv") %>%
  mutate(AFCL = AFCL*10,
         LCW = LCW*10,
         DCW = DCW*10,
         LiDe = ifelse(Status == "Li", "Live", "Dead")) %>%
  filter(S.F == "S") %>%
  group_by(Site, Year, Age, Plot) %>%
  select(LiFol, DeFol, Li.1hr, De.1hr, Li.10hr, De.10hr, Li.100hr, De.100hr) %>%
  summarise_all(sum) %>%
  ungroup() %>%
  mutate(sb_age = paste0(Year, Age))

aov.models = pfuel %>%
  select (-c(Year, Age)) %>%
  select(LiFol, DeFol, Li.1hr, De.1hr, Li.10hr, De.10hr, Li.100hr, De.100hr, Site, Plot, sb_age) %>%
  map(~ aov(.x ~ sb_age + Site/Plot, data = pfuel))

When the aov.models runs I generate this error:

Error in lm.fit(x, y, offset = offset, singular.ok = singular.ok, ...) : 
  NA/NaN/Inf in 'y'
In addition: Warning message:
In model.response(mf, "numeric") : NAs introduced by coercion

I haven't gotten to the Tukey test yet, as I cannot get past the aov function. Any suggestions would be GREATLY appreciated!

You can find the data here: https://www.dropbox.com/s/yb8rh860fc7fff2/CFL.csv?dl=0

Thanks!

My first red flag would be pfuel$... - I'm pretty sure you don't want to be referencing anything using pfuel$ really when aov(formula, data=.) or some similar construction could be used. — thelatemail
I'm also unclear why you do two separate select() statements. If one is excluding certain variables, but then you just select a specified list, why bother with the initial exclusion? — thelatemail
Thanks for the comment. Still getting the same error though. I have added the error to the main question. Thanks — nate-m
The select comment is a good one - thank you! Not sure, made sense at the time. — nate-m
Go down the list of all the variables in your select() call. When you reach aov(Site~ sb_age + Site/Plot, data = pfuel) & aov(sb_age ~ sb_age + Site/Plot, data = pfuel), do you still get output without error? — Z.Lin

Maxwell Joseph Maxwell Joseph · Accepted Answer · 2017-09-05T17:56:02

It may be easier to convert the data to long form, split by response, then fit models and feed the output to the HSD.test function, e.g.,

aov.models <- pfuel %>%
  select(-Year, -Age) %>%
  gather(variable, value, -sb_age, -Site, -Plot) %>%
  split(.$variable) %>%
  map(~ aov(value ~ sb_age + Site/Plot, data = .x)) %>%
  map(HSD.test, trt = 'sb_age')

I also removed one of the select() statements, as it was selecting all of the columns.

iterate dependent variable using purr and map (tidyverse)

2 Answers