Class Project: Data Curation¶

NEU466G, Dr. Harris · Class documents

Overview¶

Week 11 of the class research project: unbiased sampling and sources of variation.

By Week 11 the class has one data set, made by many people on many dendrites. Before the class compares the conditions, the data must be checked and cleaned. This is called data curation. Curation finds mistakes and differences in how people worked, so that a difference in the results comes from the tissue and not from the tracing.

What you do: check the class data for your group's dendrite, and fix what you find.
What you hand in: your saved series, uploaded to Canvas before you leave class, and the curation lab report, due in Week 12 (see Assignments).

Unbiased sampling¶

A sample is unbiased when every synapse or spine had the same chance to be counted. For example, counting every spine along a stretch of dendrite is unbiased. Counting only the spines that are easy to see is biased, because easy spines may be larger or clearer than the rest.

Sources of variation¶

The numbers from two dendrites can differ for reasons that have nothing to do with the condition. Some of these are:

Real biology: dendrites and spines differ from one another, even in the same animal.
The people: two people may call the same synapse a different category, or miss different polyribosomes.
The data: missing sections, section flaws, or parts of the series that are hard to see.

Curation reduces the second and third kinds, so the first kind can be measured.

Steps¶

To be decided

The curation steps are not written yet. They may include:

checking that every synapse has a Synapse_type and every spine has been searched for resources;
checking that every stamp name follows the project naming scheme;
a second person checking a sample of another group's calls;
choosing which stretch of each dendrite the class counts, so every dendrite is sampled the same way.

Dr. Harris decides the steps and what the curation lab report must contain.

Saving¶

Save with Ctrl+S (Mac: Cmd+S), back up your .jser file, and upload it to Canvas before you leave, as in Lab 1.1.