Data Tidying- with fMRI!

Objectives

This week we’re going to polish off our section on data manipulation by working with fMRI data. First we’ll go over a couple of key functions that are useful for rearranging data into “tidy” format. Then we will download an openly-available fMRI dataset and learn how to load and visualize fMRI data in R. Finally, we’ll apply the skills we’ve learned to combine and wrangle fMRI datasets and to summarize and visualize their results.

Key concepts

tidy data, reshaping functions (pivot_longer, pivot_wider), fMRI-related concepts including NIfTI file, anatomical image, functional image, voxel, time-point, time-series

Readings

You should read this chapter before you come to class:

In-class exercises

We will follow along with the examples given in the textbook through section 8.4. Create an R project called data-tidy, and save today’s work in an R markdown report called tidy.Rmd.

Then we will follow along with a set of exercises to learn about interacting with fMRI data in R. Download the fmri-data-wrangling.Rmd file to get started. See also the haxby01 page for more information about the dataset.

Weekly assignment

Download a modified version of the fmri-data-wrangling script here. This version does NOT require that you download the original data. Instead, download this csv file that contains summary data for each subject. Put these two files in the same folder, and follow the instructions in the script. You will start from the section cleverly labeled Start running from here. Below that section, you will find several questions that you will answer using the data and the skills you have learned so far. Respond to the questions by filling in the chunks that say FILL IN HERE.

Final project preparation: Read the paper associated with your final project. Download the data and make sure you are able to load it in to R. Your markdown file should include the code that you used to load it into R.

When you zip your files to submit your assignment, please do not include the original data files. You should include only the Rmd and html report files.

Chapter summary

This chapter focuses on the principles of tidy data and how to reshape data using tidyr. The below summary was created in collaboration with Google Gemini.

Key Concepts:

  • Tidy Data: A standard way of structuring datasets that makes them easier to manipulate and analyze.
  • Principles of Tidy Data:
    • Each variable forms a column.
    • Each observation forms a row.
    • Each type of observational unit forms a table.
  • Reshaping Data: Transforming data from a wide format (multiple variables in columns) to a long format (multiple observations in rows), or vice versa.

Key Functions from tidyr:

  • pivot_longer():

    • Converts data from wide to long format.
    • Syntax: pivot_longer(data, cols, names_to, values_to)
      • cols: Columns to pivot.
      • names_to: Name of the new column that will contain the pivoted column names.
      • values_to: Name of the new column that will contain the pivoted values.
    • Example: Transforming columns representing measurements at different time points into rows.

  • pivot_wider():

    • Converts data from long to wide format.
    • Syntax: pivot_wider(data, names_from, values_from)
      • names_from: Column whose values will become column names.
      • values_from: Column whose values will populate the new columns.
    • Example: Transforming rows representing different categories into columns.

  • separate():

    • Splits a single column into multiple columns.
    • Syntax: separate(data, col, into, sep)
      • col: Column to separate.
      • into: Vector of new column names.
      • sep: Separator character or regular expression.
    • Example: Splitting a date column into year, month, and day columns.

  • drop_na():

    • Removes rows with missing values (NA).
    • Syntax: drop_na(data, ...)
    • ...: Optional columns to check for NAs. If empty, checks all columns.

Practical Applications:

  • Preparing data for analysis and visualization.
  • Standardizing data formats for consistency.
  • Reshaping data to meet the requirements of specific functions or packages.

Key Takeaways:

  • Tidy data is essential for efficient data analysis.
  • tidyr provides powerful tools for reshaping data.
  • Understanding pivot_longer() and pivot_wider() is crucial for transforming data between wide and long formats.
  • drop_na() is useful for cleaning data.

Additional Concepts Related to Processing fMRI Data:

  • NIfTI file: A common file format for storing MRI data.
  • Anatomical image: An image that maps out the anatomical structures of the brain due to its sensitivity to differences between tissue types, e.g., gray and white matter. Typically only one anatomical image is collected, with high spatial resolution to visualize brain structures.
  • Functional image: An image that measures the BOLD (blood oxygen level dependent) signal, which is used as a proxy for brain activity. Typically many functional images are collected to capture how the BOLD signal changes over time, e.g., while a participant completes a cognitive task. Functional images are typically lower in spatial resolution than the anatomical image.
  • Voxel: The spatial unit of measurement for MRI, i.e., a volumetric (3-D) pixel. The smaller the voxel, the higher the resolution and the clearer the image.
  • Time-point: Related to the temporal unit of measurement for functional MRI. We collect many functional images, one per time-point.
  • Time-series: The ordered collection of time-points, reflecting the time-course of the BOLD signal.