class: center, middle, inverse, title-slide .title[ # Week 2:</br>Data Preparation ] .author[ ### Visual Data Analytics ] .institute[ ### University of Sydney ] --- <script type="text/x-mathjax-config"> MathJax.Hub.Config({ TeX: { Macros: { bA: "{\\mathbf{A}}", bC: "{\\mathbf{C}}", bv: "{\\mathbf{v}}", bw: "{\\mathbf{w}}", bx: "{\\mathbf{x}}", bX: "{\\mathbf{X}}", by: "{\\mathbf{y}}", bY: "{\\mathbf{Y}}", bw: "{\\mathbf{w}}", bW: "{\\mathbf{W}}", bK: "{\\mathbf{K}}", bz: "{\\mathbf{z}}", bU: "{\\mathbf{U}}", bS: "{\\mathbf{S}}", bff: "{\\mathbf{f}}", bD: "{\\mathbf{D}}", bL: "{\\mathbf{L}}", calL: "{\\mathcal{L}}", calM: "{\\mathcal{M}}", bbR: "{\\mathbb{R}}", bB: "{\\mathbf{B}}", bbb: "{\\mathbf{b}}", bLambda: "{\\boldsymbol{\\Lambda}}", bDelta: "{\\boldsymbol{\\Delta}}", bH: "{\\mathbf{H}}", bI: "{\\mathbf{I}}", biota: "{\\boldsymbol{\\iota}}", bN: "{\\mathbf{N}}", bQ: "{\\mathbf{Q}}" } } }); </script> # Outline - Data frames - Operations on data frames - Data types - Data profiling --- class: center, middle, inverse # Data frames --- # What is data? - Data can be *structured* or *unstructured* - Structured data has - Observations/ cases in rows - Variables/ features in columns - Think of structured data as something that can be put into a spreadsheet. --- # Example  Current as of January 2023 --- # US Cities and States  --- # Observations and variables - In the first table: - What are the observations? - What are the variables? - How about the second table? --- # Pandas .pull-left[ - The Python library *pandas* stores data in a basic object called a *data frame*. - It also provides functions for reading in and manipulating data frames. ] .pull-right[] --- # NBA Data ```python import pandas as pd NBA = pd.read_csv('../data/NBA1.csv') NBA ``` ``` ## Player Games Played Minutes Points Birthplace ## 0 LeBron James 1394 53142 37860 Akron, USA ## 1 Kevin Durant 974 35776 26565 Washington DC, USA ## 2 James Harden 962 33378 23915 Los Angeles, USA ## 3 Russell Westbrook 1054 36352 23781 Long Beach, USA ## 4 Chris Paul 1178 40623 21235 Winston-Salem, USA ## 5 Stephen Curry 852 29256 20843 Akron, USA ## 6 DeMar DeRozan 993 34087 20812 Compton, USA ## 7 Damian Lillard 734 26648 18149 Oakland, USA ## 8 Rudy Gay 1089 34135 17463 New York, USA ## 9 Paul George 765 25711 15713 Palmdale, USA ## 10 Kyle Lowry 1055 33750 15556 Philadelphia, USA ## 11 Giannis Antetokounmpo 686 22365 15283 Athens, Greece ## 12 Anthony Davis 629 21641 15075 Chicago, USA ## 13 Bradley Beal 668 23211 14772 St Louis, USA ## 14 Kyrie Irving 637 21756 14770 Melbourne, Australia ``` --- # Keys - A key uniquely identifies an entry into the database. - In the first table the player name can act as a key. - Player birthplace could not be a key since both James and Curry are born in Akron. - However, be careful with names as keys, since two players may have the same name. --- # Atomicity - Beware of variables that combine multiple pieces of information. - Birthplace can be broken down into city and country. - This may be useful if we want to visualise points scored by players born in USA and players born outside USA. - Variables should be *atomic* - This is also known as *parsing* the data. --- # Split ```python new = NBA.Birthplace.str.split(', ',expand = True) new ``` ``` ## 0 1 ## 0 Akron USA ## 1 Washington DC USA ## 2 Los Angeles USA ## 3 Long Beach USA ## 4 Winston-Salem USA ## 5 Akron USA ## 6 Compton USA ## 7 Oakland USA ## 8 New York USA ## 9 Palmdale USA ## 10 Philadelphia USA ## 11 Athens Greece ## 12 Chicago USA ## 13 St Louis USA ## 14 Melbourne Australia ``` --- # With original table ```python NBA["CityOfBirth"] = new[0] NBA["CountryOfBirth"] = new[1] NBA ``` ``` ## Player Games Played ... CityOfBirth CountryOfBirth ## 0 LeBron James 1394 ... Akron USA ## 1 Kevin Durant 974 ... Washington DC USA ## 2 James Harden 962 ... Los Angeles USA ## 3 Russell Westbrook 1054 ... Long Beach USA ## 4 Chris Paul 1178 ... Winston-Salem USA ## 5 Stephen Curry 852 ... Akron USA ## 6 DeMar DeRozan 993 ... Compton USA ## 7 Damian Lillard 734 ... Oakland USA ## 8 Rudy Gay 1089 ... New York USA ## 9 Paul George 765 ... Palmdale USA ## 10 Kyle Lowry 1055 ... Philadelphia USA ## 11 Giannis Antetokounmpo 686 ... Athens Greece ## 12 Anthony Davis 629 ... Chicago USA ## 13 Bradley Beal 668 ... St Louis USA ## 14 Kyrie Irving 637 ... Melbourne Australia ## ## [15 rows x 7 columns] ``` --- # Another example <table> <tr> <th>Player</th> <th>Points</th> </tr> <tr> <td>Steph Curry</td> <td>20843 for Warriors</td> </tr> <tr> <td>Kevin Durant</td> <td>17566 for Thunder, 5374 for Warriors, 3625 for Nets</td> </tr> </table> --- # Split <table> <tr> <th>Player</th> <th>Points for Warriors</th> <th>Points for Thunder</th> <th>Points for Nets</th> </tr> <tr> <td>Steph Curry</td> <td>20843</td> <td></td> <td></td> </tr> <tr> <td>Kevin Durant</td> <td>5374</td> <td>17566</td> <td>3625</td> </tr> </table> --- # Problems - Empty cells for Steph Curry - Not such a big issue (see missing data later) - What if Kevin Durant moves to another team? (He did...) - What if we want to include Giannis (only played for Bucks)? - Would need new columns and code that previously worked may break. --- # Solution <table> <tr> <th>Player</th> <th>Team</th> <th>Points</th> </tr> <tr> <td>Steph Curry</td> <td>Warriors</td> <td>20843</td> </tr> <tr> <td>Kevin Durant</td> <td>Thunder</td> <td>17566</td> </tr> <tr> <td>Kevin Durant</td> <td>Warriors</td> <td>5374</td> </tr> <tr> <td>Kevin Durant</td> <td>Nets</td> <td>3625</td> </tr> </table> --- # First Normal Form - New data entries can be added by adding rows only. - Now the player and team combined form key. - Overall the example so far is about getting the data into the *first normal form*. - For the purposes of visualisation the lesson is to think carefully about how the data is structured. --- class: inverse, middle, center # Operations on Data Frames --- # The simple machines - Hundreds of years ago, it was believed that all machines were made up of six simple machines - These include: levers, wheels, pulleys, screws, etc. - Nowadays machines are more complicated. - But this is a good metaphor for data frames --- # Simple machines of data - We will consider six "simple machines" of data frames. - Transforming - Sorting - Filtering - Group by/ aggregate - Reshaping (melting and casting) - Joining (merging) - By some combination of these we can 'munge' data frames into almost any data frame we need. --- # Transform - Create a new variable based on values of existing variables. - For example, in the NBA data frame we have games played and points. - Suppose we want to create a variable of points per game (PPG) --- # Transform in Python ```python NBA["PPG"]=NBA["Points"]/ NBA["Games Played"] NBA ``` ``` ## Player Games Played ... CountryOfBirth PPG ## 0 LeBron James 1394 ... USA 27.159254 ## 1 Kevin Durant 974 ... USA 27.274127 ## 2 James Harden 962 ... USA 24.859667 ## 3 Russell Westbrook 1054 ... USA 22.562619 ## 4 Chris Paul 1178 ... USA 18.026316 ## 5 Stephen Curry 852 ... USA 24.463615 ## 6 DeMar DeRozan 993 ... USA 20.958711 ## 7 Damian Lillard 734 ... USA 24.726158 ## 8 Rudy Gay 1089 ... USA 16.035813 ## 9 Paul George 765 ... USA 20.539869 ## 10 Kyle Lowry 1055 ... USA 14.745024 ## 11 Giannis Antetokounmpo 686 ... Greece 22.278426 ## 12 Anthony Davis 629 ... USA 23.966614 ## 13 Bradley Beal 668 ... USA 22.113772 ## 14 Kyrie Irving 637 ... Australia 23.186813 ## ## [15 rows x 8 columns] ``` How would you get points per minute? --- # Sort - Suppose we want to sort the data according to one of the variables. - We can use the `sort_values` function. - Consider that we want to sort by minutes played from smallest to largest. --- # Players by minutes ```python NBAbymin = NBA.sort_values(by = 'Minutes') NBAbymin ``` ``` ## Player Games Played ... CountryOfBirth PPG ## 12 Anthony Davis 629 ... USA 23.966614 ## 14 Kyrie Irving 637 ... Australia 23.186813 ## 11 Giannis Antetokounmpo 686 ... Greece 22.278426 ## 13 Bradley Beal 668 ... USA 22.113772 ## 9 Paul George 765 ... USA 20.539869 ## 7 Damian Lillard 734 ... USA 24.726158 ## 5 Stephen Curry 852 ... USA 24.463615 ## 2 James Harden 962 ... USA 24.859667 ## 10 Kyle Lowry 1055 ... USA 14.745024 ## 6 DeMar DeRozan 993 ... USA 20.958711 ## 8 Rudy Gay 1089 ... USA 16.035813 ## 1 Kevin Durant 974 ... USA 27.274127 ## 3 Russell Westbrook 1054 ... USA 22.562619 ## 4 Chris Paul 1178 ... USA 18.026316 ## 0 LeBron James 1394 ... USA 27.159254 ## ## [15 rows x 8 columns] ``` --- # Filter - Filtering involves selecting only some subset of the data. - There are many ways to do this - Select rows - Select columns - Select by a logical condition --- # Example - Suppose we only want to consider - Players with points per game greater than 20 - Players born in Akron - Players not born in the United States - These are all examples of logical conditions (either true or false). --- # Players with PPG above 20 ```python NBAppg20 = NBA.loc[NBA["PPG"]>20] NBAppg20 ``` ``` ## Player Games Played ... CountryOfBirth PPG ## 0 LeBron James 1394 ... USA 27.159254 ## 1 Kevin Durant 974 ... USA 27.274127 ## 2 James Harden 962 ... USA 24.859667 ## 3 Russell Westbrook 1054 ... USA 22.562619 ## 5 Stephen Curry 852 ... USA 24.463615 ## 6 DeMar DeRozan 993 ... USA 20.958711 ## 7 Damian Lillard 734 ... USA 24.726158 ## 9 Paul George 765 ... USA 20.539869 ## 11 Giannis Antetokounmpo 686 ... Greece 22.278426 ## 12 Anthony Davis 629 ... USA 23.966614 ## 13 Bradley Beal 668 ... USA 22.113772 ## 14 Kyrie Irving 637 ... Australia 23.186813 ## ## [12 rows x 8 columns] ``` --- # Players born in Akron ```python NBAAkr = NBA.loc[NBA["CityOfBirth"] == 'Akron'] NBAAkr ``` ``` ## Player Games Played Minutes ... CityOfBirth CountryOfBirth PPG ## 0 LeBron James 1394 53142 ... Akron USA 27.159254 ## 5 Stephen Curry 852 29256 ... Akron USA 24.463615 ## ## [2 rows x 8 columns] ``` Note that a single = denotes assignment, a double == denotes 'equals' in a logical statement. --- # Players born outside USA ```python NBAnonUS = NBA.loc[NBA["CountryOfBirth"] != 'USA'] NBAnonUS ``` ``` ## Player Games Played ... CountryOfBirth PPG ## 11 Giannis Antetokounmpo 686 ... Greece 22.278426 ## 14 Kyrie Irving 637 ... Australia 23.186813 ## ## [2 rows x 8 columns] ``` In general we can read ! as 'not' in Python --- # Group by / aggregate - Suppose we want to compare total points scored by players country of birth. - This requires two functions - The `groupby` function tells us the variable to group on (in this case Country of Birth). - The `agg` function tells us which variable to aggregate (in this case points) --- # Groupby/aggregate in Python ```python NBAg = NBA.groupby('CountryOfBirth').agg({'Points': 'sum'}) NBAg ``` ``` ## Points ## CountryOfBirth ## Australia 14770 ## Greece 15283 ## USA 271739 ``` Other ways to aggregate include mean, min and max. --- # Reshape - Often in order to produce the visualisation we want we need to reshape the data. - This is done using two functions - The function `melt` converts the data from wide to long. - The function `pivot` converts the data from long to wide. --- # Melting ```python NBAlong = NBA.melt(id_vars=['Player'],value_vars=['Games Played', 'Minutes', 'Points']) NBAlong ``` ``` ## Player variable value ## 0 LeBron James Games Played 1394 ## 1 Kevin Durant Games Played 974 ## 2 James Harden Games Played 962 ## 3 Russell Westbrook Games Played 1054 ## 4 Chris Paul Games Played 1178 ## 5 Stephen Curry Games Played 852 ## 6 DeMar DeRozan Games Played 993 ## 7 Damian Lillard Games Played 734 ## 8 Rudy Gay Games Played 1089 ## 9 Paul George Games Played 765 ## 10 Kyle Lowry Games Played 1055 ## 11 Giannis Antetokounmpo Games Played 686 ## 12 Anthony Davis Games Played 629 ## 13 Bradley Beal Games Played 668 ## 14 Kyrie Irving Games Played 637 ## 15 LeBron James Minutes 53142 ## 16 Kevin Durant Minutes 35776 ## 17 James Harden Minutes 33378 ## 18 Russell Westbrook Minutes 36352 ## 19 Chris Paul Minutes 40623 ## 20 Stephen Curry Minutes 29256 ## 21 DeMar DeRozan Minutes 34087 ## 22 Damian Lillard Minutes 26648 ## 23 Rudy Gay Minutes 34135 ## 24 Paul George Minutes 25711 ## 25 Kyle Lowry Minutes 33750 ## 26 Giannis Antetokounmpo Minutes 22365 ## 27 Anthony Davis Minutes 21641 ## 28 Bradley Beal Minutes 23211 ## 29 Kyrie Irving Minutes 21756 ## 30 LeBron James Points 37860 ## 31 Kevin Durant Points 26565 ## 32 James Harden Points 23915 ## 33 Russell Westbrook Points 23781 ## 34 Chris Paul Points 21235 ## 35 Stephen Curry Points 20843 ## 36 DeMar DeRozan Points 20812 ## 37 Damian Lillard Points 18149 ## 38 Rudy Gay Points 17463 ## 39 Paul George Points 15713 ## 40 Kyle Lowry Points 15556 ## 41 Giannis Antetokounmpo Points 15283 ## 42 Anthony Davis Points 15075 ## 43 Bradley Beal Points 14772 ## 44 Kyrie Irving Points 14770 ``` --- # Pivoting ```python NBAwide = NBAlong.pivot(index='Player', columns = 'variable') NBAwide ``` ``` ## value ## variable Games Played Minutes Points ## Player ## Anthony Davis 629 21641 15075 ## Bradley Beal 668 23211 14772 ## Chris Paul 1178 40623 21235 ## Damian Lillard 734 26648 18149 ## DeMar DeRozan 993 34087 20812 ## Giannis Antetokounmpo 686 22365 15283 ## James Harden 962 33378 23915 ## Kevin Durant 974 35776 26565 ## Kyle Lowry 1055 33750 15556 ## Kyrie Irving 637 21756 14770 ## LeBron James 1394 53142 37860 ## Paul George 765 25711 15713 ## Rudy Gay 1089 34135 17463 ## Russell Westbrook 1054 36352 23781 ## Stephen Curry 852 29256 20843 ``` --- # A better example ```python Sydney = pd.read_csv('../data/SydneyClimate.csv') Sydney ``` ``` ## Year M01 M02 M03 M04 M05 M06 M07 M08 M09 M10 M11 M12 ## 0 2017 NaN NaN NaN NaN NaN NaN NaN NaN NaN 24.9 24.8 28.6 ## 1 2018 28.5 28.1 27.4 26.7 22.2 17.7 19.9 19.3 21.0 21.9 25.0 27.1 ## 2 2019 29.6 27.7 26.9 25.1 22.7 18.6 19.8 19.5 22.0 24.7 27.0 28.1 ## 3 2020 29.0 27.5 25.6 24.5 20.3 18.7 18.2 19.5 22.7 24.2 26.1 26.1 ## 4 2021 27.4 26.7 25.5 24.3 20.9 18.1 18.2 20.7 22.5 24.2 23.2 26.1 ## 5 2022 27.7 26.7 24.8 23.7 20.8 18.3 17.0 19.8 20.6 22.6 24.3 NaN ``` Retrieved from Bureau of Meteorology --- # Melting ```python Sydlong = Sydney.melt(id_vars='Year').sort_values(by = ['Year','variable']) Sydlong ``` ``` ## Year variable value ## 0 2017 M01 NaN ## 6 2017 M02 NaN ## 12 2017 M03 NaN ## 18 2017 M04 NaN ## 24 2017 M05 NaN ## .. ... ... ... ## 47 2022 M08 19.8 ## 53 2022 M09 20.6 ## 59 2022 M10 22.6 ## 65 2022 M11 24.3 ## 71 2022 M12 NaN ## ## [72 rows x 3 columns] ``` --- # Plot <img src="02DataPrep_files/figure-html/unnamed-chunk-15-1.png" width="2400" style="display: block; margin: auto;" /> --- # Merge - Bring two data frames together - Similar to a VLOOKUP type function in spreadsheet programs such as Excel. - We can use `merge` to add information about State of birth for NBA players. --- # Merge ```python CitiesStates = pd.read_csv('../data/UScitiesstates.csv') NBAmerge = pd.merge(NBA,CitiesStates, left_on = 'CityOfBirth', right_on = 'City') NBAmerge ``` ``` ## Player Games Played ... City State ## 0 LeBron James 1394 ... Akron Ohio ## 1 Stephen Curry 852 ... Akron Ohio ## 2 James Harden 962 ... Los Angeles California ## 3 Russell Westbrook 1054 ... Long Beach California ## 4 Chris Paul 1178 ... Winston-Salem North Carolina ## 5 DeMar DeRozan 993 ... Compton California ## 6 Damian Lillard 734 ... Oakland California ## 7 Rudy Gay 1089 ... New York New York ## 8 Paul George 765 ... Palmdale California ## 9 Kyle Lowry 1055 ... Philadelphia Pennsylvania ## 10 Anthony Davis 629 ... Chicago Illinois ## 11 Bradley Beal 668 ... St Louis Missouri ## ## [12 rows x 10 columns] ``` --- # Different type of merge - Left: Keep all entries from first data frame - Right: Keep all entries from second data frame - Inner: Keep all entries that appear in both data frames - Outer: Keep all entries that appear in either data frame --- # Merge ```python CitiesStates = pd.read_csv('../data/UScitiesstates.csv') NBAmerge = pd.merge(NBA,CitiesStates, how = 'outer', left_on = 'CityOfBirth', right_on = 'City') NBAmerge ``` ``` ## Player Games Played ... City State ## 0 LeBron James 1394 ... Akron Ohio ## 1 Stephen Curry 852 ... Akron Ohio ## 2 Kevin Durant 974 ... NaN NaN ## 3 James Harden 962 ... Los Angeles California ## 4 Russell Westbrook 1054 ... Long Beach California ## 5 Chris Paul 1178 ... Winston-Salem North Carolina ## 6 DeMar DeRozan 993 ... Compton California ## 7 Damian Lillard 734 ... Oakland California ## 8 Rudy Gay 1089 ... New York New York ## 9 Paul George 765 ... Palmdale California ## 10 Kyle Lowry 1055 ... Philadelphia Pennsylvania ## 11 Giannis Antetokounmpo 686 ... NaN NaN ## 12 Anthony Davis 629 ... Chicago Illinois ## 13 Bradley Beal 668 ... St Louis Missouri ## 14 Kyrie Irving 637 ... NaN NaN ## ## [15 rows x 10 columns] ``` --- # Putting them together - In your own time, construct data frames for the following: - Each observation is a state and with the maximum points per minute (PPM) by a player from each state. - The same as above with states ranked from highest to lowest according to the maximum PPM. - The same as above but with an extra column with the player name of the player with the highest PPM in each state. - There may be more than one correct answer. --- class: inverse, middle, center # Data types --- # Data Types - Each variable measures a certain characteristic. - Characteristics can be measured in different ways - This leads to *data type* which are important for understanding - How we can transform data. - The correct visualisation to use. --- # Scales of measurement - Any old (or new) statistics textbook will introduce four scales of measurement - Nominal - Ordinal - Interval - Ratio - These are still useful (with some caveats). --- # Nominal data - Tells us something about a characteristic but there is no notion of having more or less of a characteristic. - Example: Country of birth. - Can you think of other examples? - Even if we assign numbers to nominal categories, it does not make sense to find means medians etc. - The mode still makes sense. --- # Ordinal data - Tells us whether we have more or less of a characteristic, but not how much more or less. - Example: rate players as good, very good, excellent. - If we assign numbers to nominal categories it still does not make sense to add or subtract these numbers. - However the median (and the mode) still make sense. --- # Interval/Ratio data - All numerical data is either interval or ratio data. - The differences between the two concern whether the zero point of the scale truly represents an absence of the characteristic being measured. - Best understood with an example. --- # Points per game - Suppose I constructed a new index for points per game (PPG) where a PPG of 20 now becomes a PPG of 0. - Paul George (PPG: 20.5) would have a "new" PPG of 0.5. - Steph Curry (PPG: 24.5) would have a "new" PPG of 4.5. - Does this mean that Curry is scoring 9 times as much as George? No. --- # The textbook example - The famous example is temperature. - The Celsius scale attaches 0 and 100 to the freezing and boiling point of water. - This is arbitrary - For the Kelvin scale, zero is true zero since it is a temperature where atoms have no energy (loosely speaking). --- # Does it matter? - Not that much (outside of science). - Most data we see in business are ratio data. - In general, I will use *numeric data* and *ratio data* interchangeably. - Just think carefully when dividing with numerical variables. --- # Summary .center[ <img src="img/dataops.png" width="500" height="500" /> ] --- # Some exceptions - Nominal data with two categories. - Born in US assigned 1, born outside US assigned 0. - Arithmetic mean is then the *proportion* born in US. - Likert (customer satisfaction) scales: - Strongly disagree = 1, Disagree = 2, etc. - Using an arithmetic mean is controversial but common in practice. - Time is very unusual since calendar effects are important in business. --- # Types in Pandas ```python NBA.dtypes ``` ``` ## Player object ## Games Played int64 ## Minutes int64 ## Points int64 ## Birthplace object ## CityOfBirth object ## CountryOfBirth object ## PPG float64 ## dtype: object ``` Object is text, int64 is an integer and float64 is a real number. --- class: inverse, middle, center # Data Profiling --- # Some issues - Duplicated entries - Can be removed using `drop_duplicates` function in pandas. - Data entry errors - For example, one cannot score 20000 points in 3 minutes. - Requires domain knowledge. - These issues can be discovered during visualisation. --- # Standardisations - Steph Curry may appear elsewhere in the data as - "Stephen Curry" - "Wardell Stephen Curry II" (his full name). - Similar things happen with company names Facebook/ Meta, General Motors/ GM etc. - Requires domain knowledge and some extra coding. --- # Missing Data - There are often missing data - Need to think of a good strategy to encode missing data. - In Python there is `NaN` for this. - Often will replace with a number but this is a bad strategy --- # Do not use "0" for "missing" - Suppose if games played is missing and we replace with zero. - The zeros will distort the mean. - If we compute points per game there will be division by zero. - Do not replace missing values by zeros. --- # Do not use "-999" for "missing" - Sometimes a completely implausible number such as -999 is used to denote missing. - This can lead to strange visualisations - The following example is for the Sydney temperature data --- # Temperature <img src="02DataPrep_files/figure-html/unnamed-chunk-19-3.png" width="2400" style="display: block; margin: auto;" /> --- # Dealing with missing data - Only use complete cases - Impute missing values - With a random value - With mean/median or mode - More complicated models - Report/Visualise missing data - By reporting missing data this gives a better idea of uncertainty. --- # Report missing  Source: [Guardian](https://www.theguardian.com/australia-news/commentisfree/2022/dec/13/as-2022-comes-to-a-close-australians-can-enjoy-a-reprieve-from-the-algorithms-of-fear-and-outrage) --- class: middle, center, inverse # Wrap-up --- # Conclusions - Our focus from now on will be visualisation. - Writing code to visualise messy data is hard. - Spend the time to clean your data. - Focus on the *principles* since these work for Excel, Tableau, R, Python, etc. - After you appreciate the principles, practice your coding. --- class: middle, center, inverse # Questions