What does this print?
\nprices = {\"Desk Mat XL\": 29, \"Laptop Stand\": 49}\nprices[\"Desk Mat XL\"] = prices.get(\"Desk Mat XL\", 0) + 1\nprint(len(prices), prices[\"Desk Mat XL\"])2 30","why":"Correct. The key already exists, so .get() returns 29, one is added, and the dictionary still has two keys."},{"text":"3 30","why":"Assigning to an existing key updates it; it does not add a second entry with the same name."},{"text":"2 1","why":".get() returns the stored value, 29, because the key exists. The default 0 is only used for missing keys."},{"text":"A KeyError","why":"Every key the code reads with square brackets exists, and .get() never raises for a missing key anyway."}],"answer":0},{"q":"A colleague’s function returns 3.3333333333333335 for average([10, 20, 30]) instead of 20.0. What is the bug?
def average(values):\n total = 0\n for v in values:\n total += v\n return total / len(values)total should start at 1.","why":"Starting at 1 would make every average slightly too high; it would not explain a result based on the first value only."},{"text":"len(values) counts characters instead of items.","why":"len of a list counts its items, here 3."},{"text":"The return is indented inside the loop, so the function returns after adding the first value.","why":"Correct. On the first pass total is 10, and the function immediately returns 10 / 3. Moving return back to the function’s level fixes it."},{"text":"Python cannot divide integers, so it returns the first value.","why":"Python divides integers with / and returns a float."}],"answer":2},{"q":"Which f-string prints Share: 12.5% when share = 0.125?
f"Share: {share:.1f}%"","why":"This formats 0.125 with one decimal, giving 0.1%. The f spec does not multiply by 100."},{"text":"f"Share: {share:%}"","why":"The % spec multiplies by 100 but defaults to six decimals: 12.500000%."},{"text":"f"Share: {share:,.2f}%"","why":"This prints 0.12%: two decimals of the raw fraction."},{"text":"f"Share: {share:.1%}"","why":"Correct. The % spec multiplies by 100 and adds the percent sign; .1 keeps one decimal."}],"answer":3},{"q":"A loop over rows stops with TypeError: unsupported operand type(s) for +: 'int' and 'str' on the line total = total + row["quantity"]. What is the most likely fix?
total to something else.","why":"The name is fine; the error is about the types of the two values being added."},{"text":"Convert the quantity first: total = total + int(row["quantity"]).","why":"Correct. The quantity arrived as text, and Python never adds text to a number implicitly."},{"text":"Wrap the line in try / except TypeError and skip the row.","why":"Every quantity arrived as text, so every row would be skipped and the total would stay 0, with no error to warn you."},{"text":"Start total at "" instead of 0.","why":"That would concatenate the quantities as text, giving something like "121" instead of a number."}],"answer":1},{"q":"totals = [65.0, 18.0, 72.0]. What does [t > 50 for t in totals] produce?
[True, False, True]","why":"Correct. The expression before for is a comparison, so each item becomes a boolean."},{"text":"[65.0, 72.0]","why":"That is the filtering form, [t for t in totals if t > 50], where the condition goes after the for."},{"text":"2","why":"Counting needs sum(t > 50 for t in totals); a list comprehension always builds a list."},{"text":"A SyntaxError","why":"Any expression can go before for, including a comparison."}],"answer":0}],"s-pda-2":[{"q":"After orders = pd.read_csv("orders.csv"), orders["order_date"].dt.month raises AttributeError: Can only use .dt accessor with datetimelike values. What is the fix?
orders["order_date"].month instead.","why":"Datetime parts on a column always go through .dt; without it you get another AttributeError."},{"text":"Sort the column first.","why":"Sorting text dates does not change their type."},{"text":"Load with parse_dates=["order_date"] (or convert with pd.to_datetime) so the column is datetime64.","why":"Correct. The column was read as text (object), and .dt only works on datetime columns."},{"text":"Drop the rows with missing dates.","why":"There are no missing dates; the problem is the column’s type."}],"answer":2},{"q":"What does products.loc[products["unit_price"] > 100, "name"] return?
"name" picks one column, so the result is a Series."},{"text":"True or False for each product","why":"That is the mask itself, products["unit_price"] > 100, before it is used to select."},{"text":"The number of products priced above 100","why":"Counting would need .sum() on the mask or len() on the result."}],"answer":1},{"q":"Which line raises the price of active products by 5% and reliably updates products itself?
products[products["is_active"] == 1]["unit_price"] *= 1.05","why":"Chained assignment: the filter creates a new table, and the change is applied to that copy only."},{"text":"products["unit_price"][products["is_active"] == 1] *= 1.05","why":"Also chained; under Copy-on-Write it never reaches products."},{"text":"products.unit_price * 1.05","why":"This computes new prices for every product and throws the result away."},{"text":"products.loc[products["is_active"] == 1, "unit_price"] *= 1.05","why":"Correct. One .loc call names the rows and the column, so the assignment happens on products."}],"answer":3},{"q":"What is the result of s1 + s2?
s1 = pd.Series([10, 20], index=[\"a\", \"b\"])\ns2 = pd.Series([1, 2], index=[\"b\", \"c\"])a: NaN, b: 21, c: NaN","why":"Correct. Values are matched by label; only b exists on both sides, and labels on one side only give NaN."},{"text":"a: 11, b: 22","why":"That is position-based addition, which pandas does not do for Series with different labels."},{"text":"a: 10, b: 21, c: 2","why":"Missing partners are not treated as zero; use s1.add(s2, fill_value=0) for that behaviour."},{"text":"A ValueError because the indexes differ","why":"pandas aligns differing indexes instead of raising."}],"answer":0},{"q":"You want every order from the web or the mobile app. Which filter is the clearest correct choice?
","options":[{"text":"orders[orders["channel"] == "web" or orders["channel"] == "mobile_app"]","why":"or needs one True/False, so pandas raises ValueError: The truth value of a Series is ambiguous."},{"text":"orders[orders["channel"] == ["web", "mobile_app"]]","why":"Comparing a column with a list does element-wise comparison against a list of a different length, which raises an error."},{"text":"orders[orders["channel"].isin(["web", "mobile_app"])]","why":"Correct. isin checks each value against the list and reads exactly like the question."},{"text":"orders.loc["web":"mobile_app"]","why":"That slices row labels, not channel values; the index here is the row number."}],"answer":2}],"s-pda-3":[{"q":"A quantity column loads as float64 and shows values like 1.0 and 2.0. What explains it, and what keeps the column as whole numbers?
round().","why":"Quantities here are whole numbers; the type comes from blanks, not fractions."},{"text":"Some cells are blank, classic NaN is a float, and astype("Int64") keeps integers with <NA> for the gaps.","why":"Correct. The nullable Int64 dtype was made for exactly this case."},{"text":"read_csv always loads numbers as floats; use astype(int).","why":"Columns without gaps load as int64; and astype(int) fails while NaN values are present."},{"text":"The file is corrupted; reload it with keep_default_na=False.","why":"That keeps blanks as empty strings, turning the whole column into text."}],"answer":1},{"q":"pd.to_numeric(cleaned, errors="coerce") produced 12 new NaN values. What should you do next?
fillna(0) so totals are complete.","why":"That turns 12 unknown values into real-looking zeros and understates revenue."},{"text":"Drop the 12 rows.","why":"They may be real sales in an unexpected format; dropping them loses revenue without knowing why."},{"text":"Switch to astype(float), which is stricter.","why":"It would raise on the first bad value, which tells you less than listing them all."},{"text":"Print the original strings for those rows and extend the cleaning rule for each format you find.","why":"Correct. Coercion is a way to find problem values, and the raw strings tell you what rule is missing."}],"answer":3},{"q":"What does pd.to_datetime(["03/04/2025", "15/04/2025"], format="%d/%m/%Y") return?
%m/%d/%Y, which fails on month 15."},{"text":"4 March 2025 and 15 April 2025","why":"With an explicit format every value is read the same way; pandas does not switch per value."},{"text":"Two NaT values","why":"NaT appears only with errors="coerce" for values that do not match; both values match."}],"answer":0},{"q":"Your cleaning function removes duplicates first and strips spaces from names afterwards. What can go wrong?
","options":[{"text":"Nothing; the order of steps does not matter.","why":"Duplicate detection compares exact values, so order matters."},{"text":"strip() will remove letters from the names.","why":"strip() removes only leading and trailing whitespace."},{"text":"Copies that differ only by spaces ("Ali Evans" and "Ali Evans ") are not recognised as duplicates and both survive.","why":"Correct. Clean the values first, then look for duplicates."},{"text":"drop_duplicates will raise an error on text columns.","why":"It works on any column type."}],"answer":2},{"q":"After df["country"] = df["country"].str.title(), nunique() still reports 9 countries instead of 8. What is the most likely cause?
title() does not work on capitals such as EGYPT.","why":""EGYPT".title() gives "Egypt"; case is handled."},{"text":"Some values have leading or trailing spaces, so " Egypt" and "Egypt" are still different.","why":"Correct. str.strip() must come before (or with) the case fix."},{"text":"nunique() counts missing values as a country.","why":"nunique() ignores missing values by default."},{"text":"pandas caches the old values until the file is reloaded.","why":"Assignment replaces the column immediately; nothing is cached."}],"answer":1}],"s-pda-4":[{"q":"You left-merge order_items (3,991 rows) with returns on order_item_id with indicator=True. 198 rows are both. What do the left_only rows represent?
right_only, and a left join never keeps them anyway."},{"text":"Order lines that were returned twice","why":"A line returned twice would produce two both rows, not a left_only one."},{"text":"Rows the merge failed to process","why":"Merges do not fail silently; left_only is a normal, meaningful outcome."},{"text":"The 3,793 order lines that were never returned","why":"Correct. Every item stays, and those without a matching return are marked left_only."}],"answer":3},{"q":"sales.pivot_table(index="channel", columns="year", values="revenue") returns numbers far smaller than the revenue totals. Why?
aggfunc is "mean", so each cell is the average line revenue, not the sum.","why":"Correct. Pass aggfunc="sum" for totals; always state the aggregation explicitly."},{"text":"pivot_table drops cancelled orders automatically.","why":"It never filters by status; it only aggregates what it is given."},{"text":"The values are in thousands of dollars.","why":"pandas never rescales values on its own."},{"text":"Rows with missing years were dropped.","why":"Dropping a few rows could not shrink every cell this much, and there are no missing years."}],"answer":0},{"q":"orders.resample("ME")["revenue"].sum() raises TypeError: Only valid with DatetimeIndex, TimedeltaIndex or PeriodIndex. What is the fix?
"M" instead of "ME".","why":"The alias is fine; "M" is the deprecated spelling and would hit the same error."},{"text":"Sort the orders by date first.","why":"Sorting does not give the table a datetime index."},{"text":"Put the dates in the index with set_index("order_date"), or pass on="order_date" to resample.","why":"Correct. resample needs to know which datetime values to bucket."},{"text":"Convert revenue to integers.","why":"The error is about the index, not the values being summed."}],"answer":2},{"q":"Merchandising wants products split into five groups with the same number of products in each, by price. Which call fits?
","options":[{"text":"pd.cut(products["unit_price"], 5)","why":"This makes five ranges of equal width; with a few expensive products, most items land in the first group."},{"text":"pd.qcut(products["unit_price"], 5)","why":"Correct. qcut cuts at quantiles, so each group holds about a fifth of the products."},{"text":"products["unit_price"].rank(5)","why":"rank numbers rows; its first argument is not a number of groups."},{"text":"products.groupby("unit_price").ngroups","why":"This counts distinct prices; it does not assign products to groups."}],"answer":1},{"q":"Which call returns revenue per country with country as an ordinary column rather than the index?
sales.groupby("country")["revenue"].sum()","why":"This returns a Series indexed by country."},{"text":"sales.groupby("country").sum("revenue")","why":"This sums every numeric column and still indexes by country."},{"text":"sales.pivot(index="country", values="revenue")","why":"pivot needs a columns argument and never aggregates, so it fails here, and would still fail on repeated countries."},{"text":"sales.groupby("country", as_index=False)["revenue"].sum()","why":"Correct. as_index=False keeps the group key as a column, ready for merging or exporting."}],"answer":3}],"s-pda-5":[{"q":"Web sessions view a mean of 4.2 pages and a median of 3. A product manager asks for the “typical” session. What do you report?
","options":[{"text":"4.2 pages, because the mean uses every session","why":"The mean is pulled up by a few long sessions, so it describes no typical visitor."},{"text":"Both numbers averaged, 3.6 pages","why":"Averaging two statistics produces a number with no meaning."},{"text":"3 pages, the median, ideally with a range such as the quartiles","why":"Correct. In skewed data the median is the typical value, and quartiles show the spread."},{"text":"The maximum, to show what is possible","why":"The maximum describes one extreme session, not a typical one."}],"answer":2},{"q":"added_to_cart and purchased have a correlation of 0.62 across sessions. Which conclusion is sound?
added_to_cart, a different number."},{"text":"The two columns are unrelated.","why":"0.62 indicates a fairly strong positive relationship."}],"answer":0},{"q":"Month A to month B: revenue +20%, orders +20%, average order value unchanged. What drove the growth?
","options":[{"text":"Bigger baskets","why":"Average order value did not change, so basket size cannot explain anything here."},{"text":"Higher prices","why":"Higher prices would show up as a higher average order value, which stayed flat."},{"text":"It cannot be determined without customer data.","why":"The decomposition already answers it at this level; customer data would split it further."},{"text":"More orders at the same average size","why":"Correct. Revenue = orders × average order value, and only orders moved."}],"answer":3},{"q":"You need to show eight quarters of revenue so the board can compare their sizes. Which chart?
","options":[{"text":"A pie chart with eight slices","why":"Pies hide time order and make eight similar slices hard to compare."},{"text":"Vertical bars in time order, starting at zero","why":"Correct. Bars compare amounts, time order shows the trend, and a zero baseline keeps lengths honest."},{"text":"A line chart with the y-axis starting at the smallest quarter","why":"Zooming hides how big each quarter is; for comparing amounts, start at zero."},{"text":"A histogram of quarterly revenue","why":"A histogram shows how values are distributed, losing which quarter is which."}],"answer":1},{"q":"Restart and Run All stops with NameError: name 'clean_orders' is not defined, although the notebook worked an hour ago. What happened?
clean_orders was deleted or moved below the cell that uses it; the function only existed in the old session.","why":"Correct. Hidden state hid the problem; define functions before using them, or import them from a .py file."},{"text":"pandas was upgraded during the hour.","why":"An upgrade would not make your own function name undefined."},{"text":"The CSV file was renamed.","why":"A missing file raises FileNotFoundError, not NameError."},{"text":"Restart and Run All does not support functions.","why":"It runs every cell in order, including function definitions."}],"answer":0}]},"final":[{"q":"What does sum(float(v.replace(",", "")) for v in ["1,234.50", "99.00"]) return?
"1,234.5099.00"","why":"The values are converted to floats before summing, so no strings are joined."},{"text":"A ValueError","why":"The commas are removed before float() sees the text, so both conversions succeed."},{"text":"1333.5","why":"Correct. Each string is cleaned and converted, then the floats are added."},{"text":"1234.5","why":"The generator covers both values, so 99.0 is added too."}],"answer":2},{"q":"cancelled = orders[orders["status"] == "cancelled"] on Cartwheel’s orders. What is cancelled.shape?
(2066, 7)","why":"A filter keeps only matching rows; the full table has 2,066."},{"text":"(105, 7)","why":"Correct. 105 orders were cancelled, and filtering rows keeps all 7 columns."},{"text":"(105,)","why":"That would be the shape of a Series; filtering a DataFrame returns a DataFrame."},{"text":"(7, 105)","why":"Shape is always rows first, then columns."}],"answer":1},{"q":"messy["total"].sum() returns a long string instead of a number. Which first step diagnoses it?
messy["total"].round(2)","why":"Rounding text is not possible; it raises a TypeError rather than explaining anything."},{"text":"messy.drop_duplicates()","why":"Duplicates change the size of a sum, not its type."},{"text":"messy["total"].fillna(0)","why":"Filling blanks does not turn text into numbers."},{"text":"messy["total"].dtype and a look at a few values, which reveal text such as "$65.00"","why":"Correct. An object dtype on a number column means some values contain non-numeric characters."}],"answer":3},{"q":"Support tickets can reference an order, and one order can have several tickets. Which validate fits orders.merge(tickets, on="order_id")?
"one_to_many"","why":"Correct. Each order id appears once in orders and possibly several times in tickets."},{"text":""many_to_one"","why":"That asserts the right-hand keys are unique; ticket order ids repeat, so it would raise MergeError."},{"text":""one_to_one"","why":"Ticket order ids repeat, so this raises."},{"text":"No validation is possible for this pair of tables.","why":"Validation works for any merge; you choose the relationship you expect."}],"answer":0},{"q":"You need the number of orders per customer, but only order_items (one row per line) is at hand. Which aggregation is right?
items.groupby("customer_id").size()","why":"order_items has no customer_id, and size() on lines would count lines, not orders."},{"text":"items.groupby("order_id").size()","why":"This counts lines per order, not orders per customer."},{"text":"Merge customer_id from orders, then groupby("customer_id")["order_id"].nunique()","why":"Correct. Distinct order ids per customer count orders, however many lines each has."},{"text":"items["order_id"].value_counts()","why":"This is lines per order again."}],"answer":2},{"q":"monthly is revenue per month for 2024 and 2025. Which expression compares each month with the same month a year earlier?
monthly.pct_change()","why":"This compares each month with the one before it, mixing in seasonality."},{"text":"monthly.pct_change(12)","why":"Correct. Twelve periods back is the same month last year."},{"text":"monthly.rolling(12).mean()","why":"A rolling mean smooths the series; it does not compare periods."},{"text":"monthly.diff()","why":"diff gives month-on-month changes in dollars."}],"answer":1},{"q":"What label does pd.cut([100], bins=[0, 100, 200]) give to the value 100?
(100, 200]","why":"Bins are closed on the right by default, so 100 belongs to the bin that ends at 100."},{"text":"NaN","why":"100 lies inside the edges, so it is labelled."},{"text":"[100, 200)","why":"A left-closed bin like that needs right=False; the default closes bins on the right."},{"text":"(0, 100]","why":"Correct. With the default right=True, each bin includes its right edge."}],"answer":3},{"q":"One order of 50,000 dollars is added by mistake to a table of 2,000 normal orders. Which statistic barely changes?
","options":[{"text":"The median","why":"Correct. One extreme value moves the middle position by at most one row, so the median hardly moves."},{"text":"The mean","why":"The mean rises by about 25 dollars, a large shift for a typical order of a few hundred."},{"text":"The standard deviation","why":"A single extreme value inflates the standard deviation considerably."},{"text":"The maximum","why":"The maximum jumps straight to the mistaken 50,000-dollar order."}],"answer":0},{"q":"q4 = orders[in_q4] followed by q4["flag"] = True. What happens to orders under Copy-on-Write?
orders gains a flag column with True for Q4 rows.","why":"A filtered table behaves as a copy; changes to it never flow back."},{"text":"pandas raises an error because q4 is a view.","why":"Assigning a new column to q4 is allowed; it just affects q4 only."},{"text":"Nothing; only q4 gets the new column.","why":"Correct. To mark rows inside orders, use orders.loc[in_q4, "flag"] = True."},{"text":"orders gets a flag column filled with True for every row.","why":"Nothing is written to orders at all."}],"answer":2},{"q":"df["order_date"].str.contains("2025") works, but df["order_date"].dt.year fails. What does that tell you?
.dt.year, with or without a time zone."},{"text":"The column is still text and needs pd.to_datetime before date operations.","why":"Correct. .str working and .dt failing is the signature of dates stored as strings."},{"text":"The dates are ambiguous.","why":"Ambiguity affects how text is parsed; it does not stop a datetime column from using .dt."},{"text":"The column has missing values.","why":"Missing values become NaT in a datetime column and do not break .dt."}],"answer":1},{"q":"A quarter’s revenue rose 30%. Active customers rose 30%; orders per customer and average order value were flat. What is the best summary?
","options":[{"text":"Customers spent more per order.","why":"Average order value was flat."},{"text":"Customers ordered more often.","why":"Orders per customer were flat."},{"text":"Prices went up.","why":"Price rises would raise the average order value, which did not change."},{"text":"Growth came from having more active customers, each behaving as before.","why":"Correct. Revenue = customers × orders per customer × average order value, and only customers moved."}],"answer":3},{"q":"Which habit most protects a recurring monthly report from silently reporting wrong numbers when next month’s export changes format?
","options":[{"text":"A cleaning function ending in assertions about ids, dates, totals and allowed values","why":"Correct. When the format changes, an assertion stops the run with a message instead of letting bad data reach the report."},{"text":"Rounding every number to whole dollars","why":"Rounding hides small differences; it does not detect a broken input."},{"text":"Running the notebook cells in a different order each month","why":"That makes results less reproducible, not more."},{"text":"Usingerrors="coerce" everywhere so nothing crashes","why":"Coercion without checks turns format changes into silent NaNs, the exact failure you want to catch."}],"answer":0}],"glossary":[{"id":"variable","term":"Variable","definition":"A name attached to a value with =, such as unit_price = 39, so the value can be used later; the right side is computed first, then the name points at the result.","lesson":"l-pda-1-2"},{"id":"dtype","term":"dtype","definition":"The data type shared by every value in a pandas column, such as int64, float64, object (text), bool or datetime64; a wrong dtype is behind many data bugs.","lesson":"l-pda-2-1"},{"id":"dictionary","term":"Dictionary","definition":"A Python container of key: value pairs, such as a product’s name mapped to its price; you look values up by key with d[key] or d.get(key, default).","lesson":"l-pda-1-3"},{"id":"list-comprehension","term":"List comprehension","definition":"A one-line way to build a list, [expr for x in items if condition], that applies an expression to each item and optionally filters which items are kept.","lesson":"l-pda-1-4"},{"id":"f-string","term":"f-string","definition":"A string literal prefixed with f that embeds values in braces with optional formatting, such as f"{revenue:,.2f}" for 1,234.50 or f"{share:.1%}" for 12.5%.","lesson":"l-pda-1-5"},{"id":"traceback","term":"Traceback","definition":"The report Python prints when an exception stops a program; read it from the bottom, where the last line names the error type and message, then up to the failing line.","lesson":"l-pda-1-5"},{"id":"series","term":"Series","definition":"A one-dimensional pandas array of values with an index of labels; a single column of a DataFrame is a Series.","lesson":"l-pda-2-1"},{"id":"dataframe","term":"DataFrame","definition":"The pandas table: columns that are Series sharing one row index, with methods to load, select, filter, group, join and reshape data.","lesson":"l-pda-2-1"},{"id":"index-alignment","term":"Index alignment","definition":"pandas’ rule that operations between two Series or DataFrames match values by index label, not by position; labels present on only one side produce NaN.","lesson":"l-pda-2-1"},{"id":"loc-iloc","term":"loc and iloc","definition":"The two precise selectors in pandas: .loc[rows, columns] selects by label and includes the end of a slice, while .iloc selects by integer position and excludes it.","lesson":"l-pda-2-3"},{"id":"boolean-mask","term":"Boolean mask","definition":"A Series of True and False values, usually from a comparison, used to select rows with df[mask]; its sum counts matches and its mean gives their share.","lesson":"l-pda-2-4"},{"id":"chained-assignment","term":"Chained assignment","definition":"Assigning through two selections, such as df["col"][mask] = value; it may change a temporary copy instead of the DataFrame, so use one .loc call instead.","lesson":"l-pda-2-5"},{"id":"copy-on-write","term":"Copy-on-Write","definition":"pandas behaviour, default from version 3.0 and optional in 2.x, in which every selection acts as a copy, so the only way to change a DataFrame is to assign to it directly.","lesson":"l-pda-2-5"},{"id":"nan","term":"NaN","definition":"Short for not a number: the classic pandas marker for a missing value in float columns; it is not equal to anything, so test for it with isna().","lesson":"l-pda-3-1"},{"id":"nullable-dtype","term":"Nullable dtype","definition":"A pandas dtype such as Int64, Float64, boolean or string that stores missing values as pd.NA, so integer columns stay integers even with gaps.","lesson":"l-pda-3-1"},{"id":"str-accessor","term":"str accessor","definition":"The .str attribute of a text Series that applies string methods such as strip, title, replace, split and contains to every value at once.","lesson":"l-pda-3-2"},{"id":"regular-expression","term":"Regular expression","definition":"A pattern language for text in which characters like $ and . have special meanings; pass regex=False to treat them as plain characters.","lesson":"l-pda-3-3"},{"id":"ambiguous-date","term":"Ambiguous date","definition":"A date string such as 05/08/2025 that is valid under both day-first and month-first readings; only outside information can decide which one is correct.","lesson":"l-pda-3-4"},{"id":"assertion","term":"Assertion","definition":"An assert condition, message statement that does nothing when the condition holds and stops the program with the message when it fails; used to state what clean data must look like.","lesson":"l-pda-3-5"},{"id":"groupby","term":"groupby (split-apply-combine)","definition":"The pandas operation that splits rows into groups by key, applies a calculation to each group and combines the results into one table.","lesson":"l-pda-4-1"},{"id":"named-aggregation","term":"Named aggregation","definition":"A form of agg in which each keyword names an output column and its value is a (column, function) pair, such as revenue=("revenue", "sum").","lesson":"l-pda-4-1"},{"id":"merge","term":"merge","definition":"The pandas method that combines two tables side by side by matching key values; how sets the join type and validate checks the expected key relationship.","lesson":"l-pda-4-2"},{"id":"join-type","term":"Join type","definition":"The rule for unmatched keys in a merge: inner keeps keys found on both sides, left keeps every left row, right every right row, and outer keeps all keys.","lesson":"l-pda-4-2"},{"id":"long-wide","term":"Long and wide data","definition":"Two shapes for the same data: long has one row per observation, wide spreads one variable across columns; pivot_table goes from long to wide and melt back.","lesson":"l-pda-4-3"},{"id":"resample","term":"resample","definition":"A pandas method that groups rows with a datetime index into regular periods such as days, weeks (W), month ends (ME) or quarter ends (QE), including empty periods.","lesson":"l-pda-4-4"},{"id":"year-over-year","term":"Year-over-year growth","definition":"The change compared with the same period a year earlier, such as pct_change(4) on quarterly data; it removes seasonal patterns from the comparison.","lesson":"l-pda-4-4"},{"id":"qcut","term":"qcut","definition":"A pandas function that splits values into groups of about equal size at quantiles, such as quartiles, unlike cut, which uses boundaries you choose.","lesson":"l-pda-4-5"},{"id":"median","term":"Median","definition":"The middle value of a sorted column, with half the values below it; in skewed data such as order values it describes a typical value better than the mean.","lesson":"l-pda-5-1"},{"id":"correlation","term":"Correlation","definition":"A number from -1 to 1 measuring how strongly two numeric columns move together in a straight-line pattern; it signals association, not cause.","lesson":"l-pda-5-1"},{"id":"reproducible-analysis","term":"Reproducible analysis","definition":"An analysis that anyone can rerun from the raw files, top to bottom, with no manual steps, and get the same results, using parameters, functions and assertions.","lesson":"l-pda-5-4"}]}