Getting predictions from one dataset for multiple models in R

Question

I am trying to perform the following task: take a dataset, and feed it through multiple pre-designed models so that I can see how the predictions differ based on the different models. The following is what the data generally looks like:

data_1:

  Year day station1 station2 station3 hour minute avtemp
1 2020   1    1.124    1.018    0.852   00     30        0.998
2 2020   1    1.123    1.020    0.848   01      0        0.997
3 2020   1    1.119    1.013    0.842   01     30        0.991
4 2020   1    1.124    1.016    0.845   02      0        0.995
5 2020   1    1.124    1.016    0.842   02     30        0.994
6 2020   1    1.124    1.017    0.840   03      0        0.994

Then, models were generated using a separate dataset that is structured very similarly (except for the fact that they are divided by "stand", the experimental unit for which I need to have a model for each, hence the multiple models), using the following code:


models_temp <- data_2 %>% 
  group_by(stand) %>% 
  do(modeltt = lm(projectedtemp ~ avtemp, data = .)) %>% 
  ungroup()

As you can see, the independent variable in the model matches a column in data_1, so hypothetically it should read cleanly. This code then generates a dataset with two columns: one with stand, and one with the model for each stand, with a lot of data stored in a list() format for each model, as is shown here:

stand             model
trees             list(coefficients = c(`(Intercept)` = 0.66135577718....)
shrubs            list(coefficients = c(`(Intercept)` = 0.6468382809102...)

I tried to then use various versions of add_predictions, such as below, to use these models in a list to generate predictions from the dataset:


data_3 <- spread_predictions(data = data_1, models = models_temp)

Alas, I get the following error:


Error in UseMethod("predict") : 
  no applicable method for 'predict' applied to an object of class "list"

I searched StackOverflow and couldn't find specific examples of people trying to do this, at least not without having to dramatically restructure their models. Does anyone know of maybe a better function to make what I see as a relatively simple task work, a better way to structure my models/data, or a simple fix to this error I am getting? Thank you all so much in advance.

All libraries loaded are as follows, and I believe most of this stuff relies on the "modelr" package:

library(dplyr)
library(ggplot2)
library(tidyverse)
library(gridExtra)

Please specify the packages used. Not clear about the `spread_predictions` function. May be `map(models_temp$modeltt, ~ spread_predictions(data = data_1, models = .x))` — akrun, Dec 29 '20 at 19:27
It'd be helpful if you can provide your data as a `data.frame` or `tibble` definition, or just `dput` output. — andrew_reece, Dec 29 '20 at 20:07
Okay thank you specified packages now! Yeah I'm sorry for providing the data in this format, let me look up how to provide it more clearly... — Cameron, Dec 29 '20 at 22:50
@akrun that seemed to work! Fantastic. Thank you so much for the help! Do you know the best way for me to answer this question since your help was posted as a comment rather than an answer? — Cameron, Dec 29 '20 at 22:58

score 2 · Accepted Answer · answered Dec 29 '20 at 22:58

2

We can loop over the list with map and apply the function

library(purrr)
map(models_temp$modeltt, ~ spread_predictions(data = data_1, models = .x))

answered Dec 29 '20 at 22:58

akrun

874,273
37
540
662

Getting predictions from one dataset for multiple models in R

1 Answers1