1. The latest viz of the day got an interesting display of the veterinary center distribution in India's states and union territories (UT). I like the color scheme, simplicity and the layout.
    The layout has two charts: geo map and tree map. Each shows some particular information. However, there is no interaction between them. I feel that, by adding interactions between the two charts, we complement one chart with another. Viewers can get a better understanding of the distribution such as locations, ordering etc.

    So, I added the following tweaks:
    - added cross highlighting between the two charts using dashboard actions
    - added dots on the map (Otherwise, tiny UT cities and states could be ignored.)
    - added borders to states.(Otherwise, neighboring states of the same color may look like a single polygon.)
    BTW, the map in the original workbook is built using a shape file. The resulting map is different from the default OpenStreet map and Google map. For example, the shape of Jammu and Kashmir is different. The longitudes and latitudes for cities such as Puducherry do not look right. Which one is correct? I chose to use the default OpenStreet map.

    That's the tweak of the day.

    [Update] A new viz of the day uses cross highlighting, also about India: https://public.tableau.com/en-us/s/gallery/literacy-india
    Posted by
    Labels: TweakThursday
    0

    Add a comment

  2. There are more than one ways to calculate correlation between two series/vectors in Tableau.

    1.Correlation via TabPy

    I did this because I would like to know how to set up the new Python extension: TabPy.

    In Tableau 10.1, we can take advantage of TabPy. Alexander Loth has written a nice tutorial here. If you don't have GIT installed in your computer, you can just go to http://github.com/tableau/TabPy and download the package.
    Unpack the zip file and extract the "TabPy-master" folder. Follow the steps to install TabPy. It may take some time to install but the process is smooth. It took me 30 min or so. But it may depend on network and computer.

    Use the script in the calculated field to calculate the correlation. His twbx file is no longer downloadable. I reproduced the file according to the image in the tutorial. Here is the file for you to download.

    2.Correlation via Window_Corr()

    In Tableau 10.2, there is a new function Window_Corr() which calculates the correlation between two vectors of aggregated measures.
    By using the new function, I got exactly the same result as the one via TabPy. The window function has to compute along the "Customer Name" dimension.
    This example is included in the same file.

    Postscript

    The one with TabPy can't be published to Tableau Public website because it doesn't support Python or R extension. Tableau Public desktop neither.

    Note that there is a Corr() function which is for calculating the correlation between two non-aggregated values.

    Hope this helps.
    Posted by
    0

    Add a comment

  3. In the latest #MakeoverMonday project of Week11 2017, the little data set is exciting, because it is about orgasm. In spite of its small size, people created great vizzies based on the data set.

    There are two dimensions in the data: gender and sexual orientation. I found that the gender dimension is emphasized throughout most vizzies, while the other dimension is not equally considered.

    For example this one by Andy Kriebel (thanks to whom the #MakeoverMonday project is born), where the upper part is men and the lower part is women. The sexual orientation dimension is not clearly showing. I understand Andy is doing a global sorting on percentage. That is ignoring all the dimensions within. The resulting order just happens to show men on top.
    So I tweaked it a little bit, put it into the 2-dimensional grid and here is the result.
    Now viewers can easily compare the percentage horizontally (sexual orientation) and/or vertically (gender). The global ordering only provide partial information about the data. We still like to see the comparison within each dimension such as hetero men vs women, hetero men vs gay etc.

    My intention is to emphasize visualizing the data set based on the inherent dimensions.  The dimensions provide the basis and the structure for a fair comparison.

    Voila, that's the tweak of the week.

    BTW, here is my submission of this week's #MakeoverMonday project. In the viz, I try to show visually the following:
    • Global maximum and minimum
    • Comparison per gender dimension
    • Comparison per sexual orientation dimension
    Posted by
    Labels: TweakThursday
    0

    Add a comment

  4. The #MakeoverMonday data set of  week11 2017 consists of a very small amount of data. But it is very interesting and revealing. Here is the source article. It is about the likelihood or frequency that people experience orgasm in their relationship.
    The research sample size is 52,588 in which people are grouped by gender (Men,Women) and sexual orientation (Hetero, Homo, Bi).

    The observations are derived from the viz:
    -By frequency and gender, 4 clusters can be formed:
    *Heterosexual men, other men, lesbians, other women.
    -The frequencies of "other women" are substantially lower.
    -Heterosexual men & lesbians'  frequencies are noticeably higher in their respective gender.
    -Women's frequency is lower in general than men's.
    -Heterosexuals have the biggest discrepancy between men & women.
    -Homosexuals have similar frequencies regardless of gender.

    Click image below to view the viz.
    Feel free to derive your observations and share them in the comments below.
    Posted by
    Labels: MakeoverMonday
    1

    View comments

Blog Archive
Blog Archive
Loading
Dynamic Views theme. Powered by Blogger. Report Abuse.