31:T46a,{ const points = vl.markPoint({ filled: true }); const titles = JSON.stringify([ "Bad Santa", "The Oxford Murders", "Avatar", "My Baby's Daddy" ]); const labels = vl.markText({ baseline: 'bottom', dy: -5, fontSize: 9 }) .transform(vl.filter(`indexof(${titles}, datum.Title) >= 0`)) .encode( vl.text().fieldN('Title') ); return vl.layer(points, labels) .data(movies) .transform( vl.calculate('datum["Worldwide Gross"] - datum["US Gross"]').as('Worldwide Minus US Gross'), vl.filter('datum["Worldwide Minus US Gross"] > 0 && datum["US Gross"] > 0') // log transform safety! ) .encode( vl.y().fieldQ('Worldwide Minus US Gross') .title('Worldwide minus U.S. Gross (Log Scale)') .scale({ type: 'log' }).axis({ tickCount: 8, format: '$,' }), vl.x().fieldQ('US Gross') .title('U.S. Gross (Log Scale)') .scale({ type: 'log' }).axis({ tickCount: 5, format: '$,' }), vl.tooltip().fieldN('Title') ) .title('U.S. Gross vs. Worldwide Gross') .config(titleConfig) .width(500) .height(500) .render() }32:T40f,{ const order = [ 'Adventure', 'Action', 'Thriller/Suspense', 'Comedy', 'Romantic Comedy', 'Horror', 'Drama', 'Other' ]; return vl.markArea({ interpolate: 'monotone' }) .encode( vl.x().month('Month').title('Release Month'), vl.y().fieldQ('Percentage').axis(null), vl.color().fieldN('Genre').sort(order).legend(null), vl.row().fieldN('Genre').sort(order) ) .height(40) .data(movies_grouped) .transform( vl.timeUnit('month', 'Release Date').as('Month'), vl.groupby('Genre', 'Month').aggregate(vl.count().as('Count')), vl.groupby('Genre').joinaggregate(vl.sum('Count').as('Total')), vl.calculate('datum.Count / datum.Total').as('Percentage'), ) .title('Percentage of Releases per Month by Genre') .config({ view: { continuousWidth: width / 2, stroke: null }, header: { labelAngle: 0, labelAlign: 'left', labelPadding: 4, title: null }, facet: { spacing: 5 }, ...titleConfig }) .render(); }33:T4b8,{ const order = [ "Adventure", "Action", "Thriller/Suspense", "Comedy", "Romantic Comedy", "Horror", "Drama", "Other" ]; return vl .markArea({ interpolate: "monotone" }) .encode( vl.x().month("Month").title("Release Month"), vl .y() .fieldQ("Profit") .title(null) .scale({ nice: false }) .axis({ orient: "right", domain: false, tickCount: 2, labelFlush: true, format: "$,s" }), vl.color().fieldN("Genre").sort(order).legend(null), vl.row().fieldN("Genre").sort(order) ) .height(40) .data(movies_grouped) .transform( vl.timeUnit("month", "Release Date").as("Month"), vl.groupby("Genre", "Month").aggregate(vl.median("Profit").as("Profit")) ) .title("Median Film Profit, by Month and Genre") .config({ view: { continuousWidth: width / 2, stroke: null }, header: { labelAngle: 0, labelAlign: "left", labelPadding: 4, title: null }, facet: { spacing: 10 }, ...titleConfig }) .resolve({ scale: { y: "independent" } }) .render(); }34:[] 2f:{"title":"A2 Example: Movies Data","cells":[{"id":0,"value":"md`# A2 Example: Movies Data`","pinned":false,"hidden":false,"mode":"ojs"},{"id":18,"value":"md`This is an example exploration report intended to give you a sense of the scope of Assignment 2. This particular report concerns a dataset about motion pictures. Note that the analysis is _far from perfect_! There are interesting follow-up questions the analysis does not pursue, and fundamental data quality and sample selection concerns that are largely overlooked. Ask yourself what you might do differently!`","pinned":false,"hidden":false,"mode":"ojs"},{"id":22,"value":"md`

\n## Dataset: Motion Pictures Data\n\nThis dataset contains statistics for a sample of 3,201 movies collected in 2010. The data includes movie titles, genres, and box office gross revenues, as well as audience (IMDB) and critic (Rotten Tomatoes) ratings. The dataset combines data from multiple sources: [Rotten Tomatoes](http://www.rottentomatoes.com/), [The Numbers](http://www.the-numbers.com/), and [IMDB](http://www.imdb.com/).\n\nThe data is available from the [vega-datasets](https://github.com/vega/vega-datasets) repository as [movies.json](${datasets['movies.json'].url}).\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":11,"value":"datasets = require('vega-datasets@2')","pinned":false,"hidden":false,"mode":"ojs"},{"id":584,"value":"md`Load the dataset into a DuckDB client so we can wrangle it further if needed:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":1299,"value":"client = {\n const data = await datasets[\"movies.json\"]();\n const c = await DuckDBClient.of({ movies_raw: data });\n\n // Parse the release date from string into a timestamp\n await c.query(`\n CREATE VIEW movies AS\n SELECT *, strptime(\"Release Date\", '%b %d %Y') as \"Parsed Release Date\"\n FROM movies_raw\n `);\n return c;\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":1334,"value":"SELECT * FROM movies","pinned":true,"hidden":false,"mode":"sql","output":"movies","database":"var:client"},{"id":25,"value":"md`

\n## Analysis Questions\n\n**What factors drive the box office success of motion pictures?**\n\nSpecific Questions:\n1. How do films fare in the U.S. vs. Worldwide markets?\n2. How do box office returns vary by genre?\n2. How do distributors strategize release dates for certain kinds of films?\n3. How do fan favorites and critical darlings relate to financial success?\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":30,"value":"md`

\n## 1. How do films fare in the U.S. vs. Worldwide markets?\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":136,"value":"$31","pinned":false,"hidden":false,"mode":"ojs"},{"id":595,"value":"md`Despite some salient differences for individual films, U.S. gross and worldwide (minus U.S.) gross have a high level of association. Going forward, let's focus on total worldwide measures.`","pinned":false,"hidden":false,"mode":"ojs"},{"id":453,"value":"md`

\n## 2. How do box office returns vary by genre?\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":628,"value":"md`First, let's sanity check the \\`Release Date\\` and \\`Major Genre\\` fields:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":37,"value":"{\n const area = vl.markArea()\n .encode(\n vl.x().year('Release Date').title('Release Year'),\n vl.y().count().title('Number of Films')\n );\n \n const labels = vl.markText({ baseline: 'alphabetical' })\n .data([\n { year: new Date(1985, 0, 1), count: 50, text: 'An uneven distribution...' },\n { year: new Date(2035, 0, 1), count: 10, text: 'Films from the future!?!' }\n ])\n .encode(\n vl.x().year('year'),\n vl.y().fieldQ('count'),\n vl.text().fieldN('text')\n );\n \n return vl.layer(area, labels)\n .data(movies)\n .title('Films per Year')\n .config(wideConfig)\n .render()\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":84,"value":"{\n const bars = vl.markBar()\n .encode(\n vl.x().count().title('Number of Films'),\n vl.y().fieldO('Major Genre').sort('-x').title(null)\n );\n \n const label1 = vl.markText({ baseline: 'alphabetical', align: 'left' })\n .data([{}])\n .encode(\n vl.x().datum(45),\n vl.y().value(225),\n vl.text().value('Fewer films could result in sampling error, let\\'s form larger groups...')\n );\n \n const label2 = vl.markText({ baseline: 'alphabetical', align: 'left' })\n .data([{}])\n .encode(\n vl.x().datum(280),\n vl.y().value(75),\n vl.text().value('What is in a \"null\" film?')\n );\n \n return vl.layer(bars, label1, label2)\n .data(movies)\n .title('Films per Major Genre')\n .config(wideConfig)\n .render()\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":632,"value":"md`Based on our observations, let's transform the data to refine the genre categories, focus on films released in the period of 1995-2010, and estimate profits:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":1336,"value":"CREATE OR REPLACE VIEW movies_enhanced AS\nWITH \nmovies_with_years AS (\n SELECT *, extract(year FROM \"Parsed Release Date\")::INT as year\n FROM movies\n),\nfiltered_movies AS (\n SELECT *, (\"Worldwide Gross\" - \"Production Budget\") as \"Profit\"\n FROM movies_with_years\n WHERE\n 1995 <= year AND year <= 2010 AND\n \"Worldwide Gross\" > 0\n),\nc AS (\n SELECT \"Major Genre\", count()::INT AS cnt\n FROM filtered_movies\n GROUP BY \"Major Genre\"\n),\nmovies_with_genre AS (\n SELECT\n CASE\n -- Black Comedy -> Comedy\n WHEN \"Major Genre\" = 'Black Comedy' THEN 'Comedy'\n -- Null or < 100 films -> Other\n WHEN \"Major Genre\" IS NULL OR (SELECT cnt FROM c WHERE \"Major Genre\" = m.\"Major Genre\") < 100 THEN 'Other'\n ELSE \"Major Genre\" END\n AS \"Genre\",\n *\n FROM filtered_movies AS m\n)\nSELECT *\nFROM movies_with_genre","pinned":true,"hidden":true,"mode":"sql","output":"moviesView","database":"var:client"},{"id":1123,"value":"// https://observablehq.com/@observablehq/derived-views-with-duckdb\nclient_enhanced = (moviesView, client)","pinned":true,"hidden":false,"mode":"ojs"},{"id":1343,"value":"SELECT * FROM movies_enhanced","pinned":true,"hidden":false,"mode":"sql","output":"movies_grouped","database":"var:client_enhanced"},{"id":652,"value":"md`How do the genres look now? _(Better! We have at least 100 films per group.)_`","pinned":false,"hidden":false,"mode":"ojs"},{"id":1332,"value":"SELECT \"Genre\", count(*)::INT as cnt\nFROM movies_enhanced\nGROUP BY \"Genre\"","pinned":true,"hidden":false,"mode":"sql","output":"movies_counts","database":"var:client_enhanced"},{"id":115,"value":"{\n const bars = vl.markBar()\n .encode(\n vl.x().fieldQ('cnt').title('Number of Films'),\n vl.y().fieldO('Genre').sort('-x').title(null)\n );\n \n return vl.layer(bars)\n .data(movies_counts)\n .title('Films per Grouped Genre')\n .config(wideConfig)\n .render()\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":655,"value":"md`How does profit vary by genre? Let's look at all the data points along with per-genre medians and inter-quartile ranges:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":219,"value":"{\n const points = vl.markPoint({ filled: true })\n .encode(\n vl.x().fieldQ('Profit')\n .scale({ type: 'linear' }).axis({ format: '$,d', tickCount: 5 })\n .title('Profit'),\n vl.tooltip().fieldN('Title')\n );\n\n const iqr = vl.markRect()\n .encode(\n vl.x().q1('Profit'),\n vl.x2().q3('Profit'),\n vl.color().value('lightgray')\n );\n \n const median = vl.markRule()\n .encode(\n vl.x().median('Profit'),\n vl.color().value('red')\n );\n \n return vl.layer(iqr, points, median)\n .facet({\n row: vl.row().field('Genre')\n .sort(vl.median('Profit').order('descending'))\n })\n .data(movies_grouped)\n .title('Profit by Grouped Genre')\n .config({\n view: { continuousWidth: width - 100, stroke: null },\n header: { labelAngle: 0, labelAlign: 'left', labelPadding: 4, title: null },\n facet: { spacing: -1 },\n ...titleConfig\n })\n .render()\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":688,"value":"md`Filtering out the three big outliers making more than $1B:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":684,"value":"{\n const points = vl.markPoint({ filled: true })\n .encode(\n vl.x().fieldQ('Profit')\n .scale({ type: 'linear' }).axis({ format: '$s', tickCount: 10 })\n .title('Profit'),\n vl.tooltip().fieldN('Title')\n );\n\n const iqr = vl.markRect()\n .encode(\n vl.x().q1('Profit'),\n vl.x2().q3('Profit'),\n vl.color().value('lightgray')\n );\n \n const median = vl.markRule()\n .encode(\n vl.x().median('Profit'),\n vl.color().value('red')\n );\n \n return vl.layer(iqr, points, median)\n .facet({\n row: vl.row().field('Genre')\n .sort(vl.median('Profit').order('descending'))\n })\n .data(movies_grouped)\n .transform(vl.filter('datum.Profit < 1e9'))\n .title('Profit (Less than $1B) by Grouped Genre')\n .config({\n view: { continuousWidth: width - 100, stroke: null },\n header: { labelAngle: 0, labelAlign: 'left', labelPadding: 4, title: null },\n facet: { spacing: -1 },\n ...titleConfig\n })\n .render()\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":902,"value":"md`For most genres, about 75% of films make at least some profit. For the Action and Adventure films in our sample, at least 25% make over $100m!`","pinned":false,"hidden":false,"mode":"ojs"},{"id":706,"value":"md`

\n## 3. How do distributors strategize release dates for certain kinds of films?\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":775,"value":"md`Let's look at when in the year films from different genres tend to be released:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":494,"value":"$32","pinned":false,"hidden":false,"mode":"ojs"},{"id":778,"value":"md`Adventure and Action releases peak in summer, Thrillers and Comedies peak in the Spring, Horror films around Halloween, and Dramas in advance of Oscar season...`","pinned":false,"hidden":false,"mode":"ojs"},{"id":808,"value":"md`Do profits align with these peak release months?`","pinned":false,"hidden":false,"mode":"ojs"},{"id":745,"value":"$33","pinned":false,"hidden":false,"mode":"ojs"},{"id":781,"value":"md`Somewhat. Adventure and Action are most profitable in May, Comedies in the summer or end of year. Horror profits are robust, while Other does not fare well.`","pinned":false,"hidden":false,"mode":"ojs"},{"id":456,"value":"md`

\n## 4. How do fan favorites and critical darlings relate to financial success?\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":815,"value":"md`Let's first sanity check the ratings fields:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":272,"value":"vl.hconcat(\n vl.layer(\n vl.markBar()\n .encode(\n vl.x().fieldQ('IMDB Rating').bin({ maxbins: 20 }),\n vl.y().count()\n ),\n vl.markText()\n .data([{}])\n .encode(\n vl.x().datum(3.25),\n vl.y().datum(310),\n vl.text().value('Skewed Bell-Shape Distribution')\n )\n ),\n vl.markBar()\n .data(movies_grouped)\n .transform(vl.filter('datum[\"IMDB Rating\"] == null'))\n .encode(vl.y().count().axis({ orient: 'right' }).title('Count of Null Values')),\n )\n .data(movies_grouped)\n .resolve({ scale: { y: 'shared' } })\n .title('IMDB Ratings')\n .config(titleConfig)\n .render()","pinned":false,"hidden":false,"mode":"ojs"},{"id":277,"value":"vl.hconcat(\n vl.layer(\n vl.markBar()\n .encode(\n vl.x().fieldQ('Rotten Tomatoes Rating').bin({ maxbins: 20 }),\n vl.y().count()\n ),\n vl.markRule({ strokeDash: [3, 3] })\n .data([{}])\n .encode(\n vl.y().datum(120),\n vl.color().value('black')\n ),\n vl.markText()\n .data([{}])\n .encode(\n vl.x().datum(40),\n vl.y().datum(150),\n vl.text().value('Relatively Uniform Distribution')\n )\n ),\n vl.markBar()\n .transform(vl.filter('datum[\"Rotten Tomatoes Rating\"] == null'))\n .encode(vl.y().count().axis({ orient: 'right' }).title('Count of Null Values')),\n )\n .data(movies_grouped)\n .resolve({ scale: { y: 'shared' } })\n .title('Rotten Tomatoes Ratings')\n .config(titleConfig)\n .render()","pinned":false,"hidden":false,"mode":"ojs"},{"id":335,"value":"vl.layer(\n vl.markPoint({ filled: true })\n .encode(\n vl.x().fieldQ('Rotten Tomatoes Rating').title('Rotten Tomatoes Rating'),\n vl.y().fieldQ('IMDB Rating').title('IMDB Rating'),\n vl.tooltip().fieldN('Title')\n ),\n vl.markLine({ color: 'black', strokeDash: [5, 5] })\n .transform(vl.regression('IMDB Rating').on('Rotten Tomatoes Rating').as('x', 'y'))\n .encode(\n vl.x().fieldQ('x'),\n vl.y().fieldQ('y')\n )\n )\n .data(movies_grouped)\n .render()","pinned":false,"hidden":false,"mode":"ojs"},{"id":818,"value":"md`The two ratings systems are distributed differently, but both look reasonable and associate more or less as we might expect. Also, both exhibit null values.\n\nHow predictive are ratings of profits? _(Ignoring the three outliers we filtered before.)_`","pinned":false,"hidden":false,"mode":"ojs"},{"id":355,"value":"{\n const subPlot = (field, note) => vl.layer(\n vl.markPoint({ filled: true })\n .encode(\n vl.x().fieldQ(field).title(`${field} (${note})`),\n vl.y().fieldQ('Profit').title('Profit').axis({ format: '$,s' }),\n vl.tooltip().fieldN('Title')\n ),\n vl.markLine({ color: 'black', strokeDash: [5, 5] })\n .transform(vl.regression('Profit').on(field).as('x', 'y'))\n .encode(\n vl.x().fieldQ('x'),\n vl.y().fieldQ('y')\n )\n );\n\n return vl.hconcat(\n subPlot('Rotten Tomatoes Rating', 'Critics').width(350),\n subPlot('IMDB Rating', 'Viewers').width(350)\n )\n .data(movies_grouped)\n .transform(vl.filter('datum.Profit < 1e9'))\n .render();\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":834,"value":"md`We see a positive association between ratings and profits. Linear regression fits corroborate this, but do not fit the data particularly well. It appears that the more lucrative films do tend to get higher ratings, especially from IMDB viewers, though less so for Rotten Tomatoes critics. Unliked films reliably fare poorly, while plenty of well-rated films also fail to make much money.\n\nThe results are biased up by outliers. What about films making less than $200M?`","pinned":false,"hidden":false,"mode":"ojs"},{"id":857,"value":"{\n const subPlot = (field, note) => vl.layer(\n vl.markPoint({ filled: true })\n .encode(\n vl.x().fieldQ(field).title(`${field} (${note})`),\n vl.y().fieldQ('Profit').title('Profit').axis({ format: '$,s' }),\n vl.tooltip().fieldN('Title')\n ),\n vl.markLine({ color: 'black', strokeDash: [5, 5] })\n .transform(vl.regression('Profit').on(field).as('x', 'y'))\n .encode(\n vl.x().fieldQ('x'),\n vl.y().fieldQ('y')\n )\n );\n\n return vl.hconcat(\n subPlot('Rotten Tomatoes Rating', 'Critics').width(350),\n subPlot('IMDB Rating', 'Viewers').width(350)\n )\n .data(movies_grouped)\n .transform(vl.filter('datum.Profit < 2e8'))\n .render();\n}","pinned":false,"hidden":false,"mode":"ojs"},{"id":863,"value":"md`The positive association remains, but we now see more variation relative to Rotten Tomatoes Ratings. You can still make money if the critics do not like your film!`","pinned":false,"hidden":false,"mode":"ojs"},{"id":923,"value":"md`The take-away? \"Better\" films do seem to make more money on average, but neither critic's praise nor popularity provide a guarantee. While studios might strive to make great films, it seems that other, more easily controlled, factors such as genre and release schedule might lead to more reliable projections of future ticket sales.`","pinned":false,"hidden":false,"mode":"ojs"},{"id":5,"value":"md`

\n## Appendix\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":320,"value":"titleConfig = ({\n title: { anchor: 'start', frame: 'group' }\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":75,"value":"wideConfig = ({\n view: { continuousWidth: width },\n autosize: { type: 'fit-x', contains: 'padding' },\n ...titleConfig\n})","pinned":false,"hidden":false,"mode":"ojs"},{"id":1217,"value":"import {vl} from '@vega/vega-lite-api-v5'","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W34","stdlib":"1","titling":"auto"}