Train and analyze many models for #TidyTuesday crop yields | Julia Silge

[utterances-bot]

Train and analyze many models for #TidyTuesday crop yields | Julia Silge

Learn how to use tidyverse and tidymodels functions to fit and analyze many models at once.

https://juliasilge.com/blog/crop-yields/

[ntihemuka]

do you have a tutorial to check how models fits the dadta using glane?

[juliasilge]

Do you mean broom::glance()? You can check out this article on tidymodels.org, and maybe this one as well.

[ntihemuka]

great helpful articles. thanks

…

On Mon, May 24, 2021 at 3:55 PM Julia Silge ***@***.***> wrote: Do you mean broom::glance()? You can check out this article on tidymodels.org <https://www.tidymodels.org/learn/statistics/tidy-analysis/>, and maybe this one as well <https://www.tidymodels.org/learn/statistics/k-means/>. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#30 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AS3DBKI3F4WUNVGTOR54FATTPJSGZANCNFSM45M2Y5UQ> .

[MohsinRamay]

Amazing work Julia especially for the data links. Do you have any expertise over Functional Data Analysis?

[poligabi]

Hello, it worked once with my data. But when I tried again with the same data.

serial_lm<- serialdt %>%
nest(yndvi = c(year, ndvi)) %>%
mutate(model = map(yndvi, ~ lm(ndvi ~ year, data = .x)))

Keep caming this error message:

Erro: Problem with mutate() column model.
i model = map(yndvi, ~lm(ndvi ~ year, data = .x)).
x 0 (non-NA) cases
Run rlang::last_error() to see where the error occurred.

But I already checked (several times and ways) and my data have not any NA.
Is there any other reason for this type of error?

[juliasilge]

@poligabi If you have some kind of problem where it worked one time before, I would suggest restarting R and rerunning in a fresh R session to get anything really messed up out of your environment.

[conlelevn]

@juliasilge Even though this screencast have used a basic model but I still can learn new things in your data manipulation skill which its really great :)
Just one question in this: when we do EDA to visualize the crop yield of several countries, by eyes only we can see there is a obvious increase trend in overall, do you think in practical we still need to use model to confirm it?

[juliasilge]

@conlelevn HA yep, if you have an effect you can clearly see in visualization, you probably don't need to confirm it with a model. In this example, I think that using statistical modeling gives you a bit more, like which countries have the bigger/smaller increases in crop yields, and which have more uncertainty about.