Regenerate

There might be times when you need to rebuild a chart. It might be something you created and you’ve lost the original data. Or it might be someone else’s work and you want to integrate their results with your charting style.

How do you get the data values?

LLMs can process images. In some cases, they can figure out data values from simple charts.

Bar Chart Example

The following bar chart is from one of my 1982 publications (McMillan et al. 1982). The original chart was created on a pen plotter. The data are no longer available.

The goal is to recreate this chart using the specifications stored with Claude.

Prompt

I need to extract the data from this chart. Can you help?

Original Chart

Prompt

I’d like the data retrieved from this diagram. Then, using the Project knowledge guidelines, I would like a table of the data and a bar chart that is in the same general format as the original diagram. Please double check the use of the guidelines.

Response

Certainly, I’ll retrieve the data from the diagram, create a table, and then produce a bar chart following the provided guidelines. Let’s go through this step-by-step.

First, let’s extract the data from the image:

Show the code
## Set read_csv function option to suppress warning messages.
options(readr.show_col_types = FALSE)

## Use the package for the read_csv function
library(tidyverse)
library(readr)

## Data input.
data <- read_csv(col_names=TRUE, file=
            "elongation_range,    percentage
             0,                   38
             1-5,                 22
             6-10,                18
             11-15,               13
             16-20,               3
             21-25,               1
             26-30,               1")

## Write the data to a CSV file for later use
write_csv(data, "seedling_elongation_data.csv")

## Display the data
data
# A tibble: 7 × 2
  elongation_range percentage
  <chr>                 <dbl>
1 0                        38
2 1-5                      22
3 6-10                     18
4 11-15                    13
5 16-20                     3
6 21-25                     1
7 26-30                     1

Now, let’s create a table using the gt package as specified in the guidelines:

Show the code
## Load required libraries
library(gt)

## Initialize the identification information.
table_title <- "Hypocotyl-Plumule Elongation in Seedlings"
data_source <- "Image data"

## Create the table.
table <- gt(data) %>% 
  tab_header(
    title = table_title
  ) %>% 
  tab_source_note(
    source_note = paste0("Source: ", data_source)
  ) %>%
  tab_style(
    style = cell_text(v_align="top"),
    locations = cells_body()
  ) %>%
  fmt_number(
    columns = percentage,
    decimals = 0
  ) %>%
  cols_label(
    elongation_range = "Elongation Range (mm)",
    percentage = "Percentage of Seedlings"
  ) %>%
  tab_footnote(
    footnote = "mm",
    locations = cells_column_labels(columns = elongation_range)
  )

## Output the table
table
Hypocotyl-Plumule Elongation in Seedlings
Elongation Range (mm)1 Percentage of Seedlings
0 38
1-5 22
6-10 18
11-15 13
16-20 3
21-25 1
26-30 1
Source: Image data
1 mm

Now, let’s create a bar chart using ggplot2, following the guidelines provided:

Show the code
## Load required libraries
library(ggplot2)

## Set the source
source <- "Image data"

## Create the bar chart
ggplot(data, aes(x = elongation_range, y = percentage)) +
  geom_bar(stat = "identity", fill = "cornsilk3", color = "black") +
  theme_minimal() +
  theme(
    panel.background = element_rect(fill = "darkslategray2"),
    axis.text = element_text(color = "black", size = 12),
    axis.title = element_text(color = "black", size = 12),
    panel.grid.major.x = element_blank(),
    panel.grid.minor.x = element_blank()
  ) +
  scale_y_continuous(
    limits = c(0, max(data$percentage) * 1.1),
    expand = c(0, 0)
  ) +
  labs(
    x = "Hypocotyl-Plumule Elongation (mm)",
    y = "% of Seedlings",
    caption = source
  )

I’ve followed the guidelines provided in the documents to create the data input, table, and bar chart. Here’s a summary of how I adhered to the guidelines:

  1. Data Input:
    • Used the specified R chunk initialization.
    • Used read_csv for data input and write_csv to save the data.
    • Arranged the data for easy viewing.
  2. Table:
    • Used the gt() function to create the table.
    • Created variables for the table title and data source.
    • Aligned table content at the top of the cells.
    • Added units (mm) as a footnote.
    • Formatted the percentage column as integers.
  3. Bar Chart:
    • Used “cornsilk3” as the fill color for the bars.
    • Set the panel background to darkslategray2.
    • Made axis labels and scale values black with size 12.
    • Used y-axis grid lines and removed x-axis grid lines.
    • Started the y-axis at zero with a small extension at the top.
    • Added the source information as a caption.

These implementations should closely match the original diagram while adhering to the provided guidelines. If you need any modifications or have any questions, please let me know.

Observations

There are a few small modifications that would help the results.

  • Modifying the source note so it references the publication, rather than “Image data.”

  • Removing the units (mm) from the column heading in the table.

  • Adding “Cymodocea” to the title instead of just “Seedlings.”

Each of these changes is quite straightforward in the R code.

Overall, this regeneration example worked quite well.