pandas_panel#

  • Series: lecture-python.myst

  • File: lectures/pandas_panel.md

  • Audit date: 2026-08-26

  • Corpus snapshot: e25fdf2345

  • Categories audited: writing, code, figures, links, admonitions (JAX out of scope)

  • Overall score: 6.9 / 10

  • Priority: HIGH

Score breakdown#

Category

Score

One-line note

Writing

3.5/10

qe-writing-006 ×4; qe-writing-005 ×3; qe-writing-003 ×4, +3 more.

Math

N/A

no mathematical content.

Code

7.5/10

qe-code-001 ×6.

JAX

out of scope

JAX rules target lecture-jax.

Figures

4.5/10

qe-fig-006 ×5; qe-fig-005 ×8; qe-fig-003 ×4, +2 more.

References

N/A

no citations in this lecture.

Links

9/10

qe-link-002 ×1.

Admonitions

10/10

no mechanical violations detected.

Issues#

Critical#

None found.

High severity#

  • [qe-code-001] (reviewer) — Follow PEP8 unless closer to mathematical notation. Count: 6. Lines: 302, 365, 375, 528, 610, 620. Example: 302-304 loops over continent names in a variable called country: replace = ['Central America', 'North America', 'South America'] then for country in replace: - the name says country, the values are continents, and the list is named after the method being called on it. The whole loop is one call, merged.Continent.replace(replace, 'America'), and its continuation at 304 is indented to column 36 under an opening delimiter at column 48. matplotlib.pyplot and seaborn are imported at 365-366, three hundred lines after pandas at 72 and immediately before their first use, so the file has module-level imports in two places rather than one cell at the top. 371 and 375 both compute merged.mean().sort_values(ascending=False), and 375 runs to 103 characters. Comma spacing is inconsistent within single lines - columns=['UNIT','AGE', 'SEX', 'INDIC_EM', 'GEO'] at 527 has the space after four of five commas, swaplevel(0,-1) at 576, bbox_to_anchor=(1,0.5) at 615 - and 528 leaves one space before an inline comment where PEP8 asks for two. 610 writes palette=("husl"), which is a parenthesised string rather than the tuple the parentheses suggest, and plots y=0, naming the value column by the integer label reset_index() happened to give it. And the file’s last line, 620, is a closing fence with a leading space, ```.

  • [qe-fig-005] — Descriptive figure names for cross-referencing. Count: 8. Lines: 234, 364, 370, 391, 408, 418, 472, 608. Example: {figure} without :name:.

  • [qe-fig-006] — Lowercase axis labels. Count: 5. Lines: 377, 395, 412, 423, 613. Example: axis label Country.

  • [qe-writing-006] — Capitalize lecture titles properly. Count: 4. Lines: 60, 175, 344, 484. Example: H2 Title Case: ‘Slicing and Reshaping Data’ (Reshaping, Data).

Medium severity#

  • [qe-fig-003] — No matplotlib embedded titles. Count: 4. Lines: 393, 410, 421, 478. Example: plt.title.

  • [qe-link-002] — Use doc links for cross-series references. Count: 1. Lines: 34. Example: raw link to python-programming.quantecon.org.

  • [qe-writing-001] — Use one sentence per paragraph. Count: 1. Lines: 508. Example: 2 sentences in one paragraph.

  • [qe-writing-002] (reviewer) — Keep writing clear, concise, and valuable. Count: 3. Lines: 104, 161, 228. Example: 104 misspells a pandas class and misplaces a comma in the same sentence: “To more easily filter our time series data, later on, we will convert the index into a DateTimeIndex” - the class is DatetimeIndex, which is how the lecture spells it correctly at 151, 320, 328 and 336, and the parenthetical commas around “later on” detach it from the clause it modifies. 161 reads “For the rest of lecture”, dropping the article. And 228-230 loses a preposition in the sentence that explains the join the whole section turns on: “Here we will pass how='left' to keep all countries in realwage_f, but discard countries in worlddata that do not have a corresponding data entry realwage_f” - it should be “entry in realwage_f”.

  • [qe-writing-003] (reviewer) — Maintain logical flow. Count: 4. Lines: 416, 434, 465, 553. Example: 419 does more than the prose says. 416 announces “We will drop Australia as a continent for plotting purposes”, but merged = merged.drop('Australia', level='Continent', axis=1) rebinds merged for the rest of the lecture, so every later cell silently excludes Australia: the .describe() at 431, the grouped object at 449, and grouped.size() at 462 - which 456-457 introduces as “the number of countries in our dataset for each continent”, and it is no longer the dataset - and the closing kernel-density figure at 472-481, which 51 promised as “kernel density estimates of wages by continent”. Second, 465-467 says the density estimate is “of real minimum wages in 2016”, while the code at 476 selects .loc['2015'] and the title at 478 says 2015. Third, 434 (“This is a simplified way to use groupby”) sits directly under the .describe() cell at 430-432, so “This” reads as .describe(), which is not a use of groupby at all; the intended antecedent is the merged.T.groupby(...) calls twenty lines earlier at 403 and 409. Fourth, the second exercise describes its own filter three incompatible ways: 553-554 asks for “employment as a percentage of ‘active population’”, 592-593 says “percentage employed in the active population”, and 596-598 selects UNIT='Percentage of total population' with INDIC_EM='Active population' - which is the active population as a percentage of the total population, the reciprocal framing - and the figure is then titled “Employment in Europe (2015)” at 614.

  • [qe-writing-005] (reviewer) — Use bold for definitions, italic for emphasis. Count: 3. Lines: 89, 91, 436. Example: there is not one bold or italic span in the file, and three terms are introduced that the rule would put in bold. 89 introduces “long format” (“The data is currently in long format, which is difficult to analyze when there are several dimensions to the data”) and 91 its counterpart “wide format”, and these two are the axis the entire first half of the lecture turns on. 436 introduces the groupby idiom with single quotes - “Using groupby generally follows a ‘split-apply-combine’ process” - and then defines each of the three steps in the list at 438-440, so it is a definition being marked as a quotation. The same single-quote-for-a-term habit appears at 44 (‘panel’ and ‘data’), 197 (‘Country’), 242-246 (‘left’ / ‘right’ dataframe) and 240 (‘on’).

  • [qe-writing-007] (reviewer) — Use visual elements to enhance understanding. Count: 3. Lines: 232, 408, 472. Example: the figures at 408-414 and 418-425 are the same four lines of code and carry the identical title, plt.title('Average real minimum wage') at 410 and 421, differing only in whether Australia is in the frame - so two adjacent figures are indistinguishable on the page and the one sentence between them (416) is all that tells them apart. The Venn diagram at 234 is a static PNG with an empty {figure} body, introduced by “This is illustrated by the red shading in the following diagram” (232): the red region is the entire content of the figure, and nothing in the page text or a caption says which set it is, so a reader who cannot see the image gets nothing. And the kernel-density figure at 472-481 is the lecture’s headline deliverable, promised at 51, yet it is the only figure with no interpretive sentence after it - 465-467 explains how it is built and the lecture then moves to “## Final Remarks”, so the reader is shown four overlapping densities and told nothing about what they show.

Low severity#

  • [qe-fig-001] — Do not set figure size unless necessary. Count: 1. Lines: 367. Example: style override.

  • [qe-fig-002] — Prefer code-generated figures. Count: 1. Lines: 234. Example: static image .png.

Strengths#

  • Every transformation is shown one step at a time with its own cell and its own head(), so the reader can see the frame change shape: long to wide by pivot_table (97-102), the index to datetime (106-109), the MultiIndex levels named and inspected (118-124), a top-level selection (129-131), .stack() with and without an explicit level (140-149), a year selected and the two lower levels stacked into a cross-section (157-159), and finally .xs down to the hourly 2015-dollar frame (169-173).

  • The merge is set up by explaining the choice before making it: 218-224 lists all four join types in one four-item list, 226 says what merge does by default, 228-230 says which one is being used and why, and 238-246 then explains each of the two key arguments in terms of where the country name actually lives in each frame (left_index=True because it is the index on the left, right_on='Country' because it is a column on the right).

  • Missing data is handled as a teaching moment rather than swept away: 257-262 checks for it with .isnull(), 264 states the count out loud (“We have three missing values!”), 266-280 builds the repair dictionary and shows what .map() alone produces, and 282-285 then explains precisely why .fillna() is used instead of assignment - “only fills in NaN values … while leaving other values in the column unchanged” - which is the mistake a reader would otherwise make.

  • The groupby section is structured around the split-apply-combine idea and then walks the three steps in order: 436-443 names them and says which one groupby itself performs, 448-451 creates and displays the DataFrameGroupBy object without aggregating so the reader sees what it is, 453-463 applies .size() as the simplest possible aggregation, and 465-481 then applies a real one.

  • Both exercises use data of a genuinely different shape from the body’s - a five-level Eurostat MultiIndex (UNIT, AGE, SEX, INDIC_EM, GEO) - and the solutions teach transferable idioms: enumerating every level’s values with a loop over employ.columns.names (541-544), promoting a level with swaplevel and re-sorting (575-578), and filtering out non-country GEO entries with a list comprehension on the level values (585-590).

  • The lecture is honest about what it is not: 486-491 names what was covered and points at xarray for the N-dimensional case rather than implying pandas is the end of the road, and 34 links the prerequisite pandas lecture in the first sentence.

  • Both exercises are gated with {exercise-start} / {exercise-end}, both solutions carry :class: dropdown (520, 569), the second exercise’s {hint} is a dropdown with exactly the one fact needed (559-563, “GEO includes both areas and countries”), and all five admonition rules measure zero.