I didn't cover it in the video, but there's a widely used convention for determining what's an outlier and showing them in a box plot. This article describes it: www.real-statistics.com/excel-capabilities/creating-box-plot-outliers-manually/
Well, skewness in general (not just 1.113) will appear in a box plot as the "whisker" and/or "box" sections at one end of the box and whisker shape being shorter than the box and whisker shapes at the other end. In a distribution heatmap, skewness appears as colored cells at one extremity of a column of cells being darker than cells at the other extremity. Kind of hard to explain without visual aids...
Thanks, Janet. This is a bit of a coin toss, to be honest. In general, scales with intervals based on 1, 2 or 5 (integers that are naturally divisible into 10) make it easiest for people to perceive values in a chart so, yes, intervals of 30% (i.e., based on 3) aren't ideal. Intervals of 25% might be better than 30%, but I tend to avoid intervals based on 2.5 (not an integer that's naturally divisible into 10). The question, then is whether intervals of 5%, 10%, 20% or 50% would work better. Of these, only 20% would "fit" in a small scale like the one in the frequency heatmap. There would be a lot of stops on that small scale, though (0%, 20%, 40%, 60%), so it might be crowded-looking. Like I said, a bit of a coin toss between 30% (less intuitive but cleaner-looking) and 20% (more intuitive but more cluttered-looking).
It was actually created in Excel, using conditional formatting (and making the numbers in the cells invisible, see support.microsoft.com/en-au/office/hide-or-display-cell-values-c94b3493-7762-4a53-8461-fb5cd9f05c33 )
@@practicalreportingthis will only work if the value we want to represents are inside the table. What if we wanted to see the frequency of that value and represent that in the heatmap? in this vid that would be for example 20 people with >120k salary. Our data would have 20 different names with >120k salary. To represent that 20 people we need a table of countif salary>120k and we need to do this for each salary band. Then we do the conditional formatting.. CMIIW. Good vid!!!
@marior6662 To make a distribution heatmap, yes, each cell has to contain (or be associated with) the number (or %) of values that fall within that cell. That value then determines the color of that cell. The example that I showed was created in Excel using conditional formatting, and the numbers in the cells were made invisible using this trick: support.microsoft.com/en-us/office/hide-or-display-cell-values-c94b3493-7762-4a53-8461-fb5cd9f05c33#:~:text=Hide%20cell%20values&text=On%20the%20Home%20tab%2C%20click,Type%20%3B%3B%3B%20(three%20semicolons).
Hi Nick, congrats on this exemplary explanation about box plot. Curious to know what´s your opinion on why people don´t find this chart very intuitive? (minute 04:20) .... are there any like sociological or maybe anthropological reasons to explain why people have a hard time understanding a very informative plot? Please if you have any links to refer me, I would greatly appreciate it. Lastly, do you make those charts on the video with Excel or Tableau? thanks, mate, greetings from Perú.
Glad that you found the video to be useful. My comment about many people finding box plots to be unintuitive is based on my experiences explaining them to thousands of workshop participants, and the fact that they require an understanding of the abstract notion of quartiles (which very few people possess in most organizations) in order to be interpreted. I suspect that many people find frequency heatmaps to be more intuitive since we pre-attentively associate higher color intensity with higher quantities (in this case, higher concentrations of values), whereas people have weaker pre-attentive associations for box and whisker shapes. Frequency heatmaps also only require an understanding of bins, which are easier to grasp than the concept of quartiles. If you have a statistical background, you may find these reasons to be kind of silly (i.e., "quartiles aren't that hard to understand..."), but most people don't have statistical backgrounds ;-)
Hi Nick, thank you so much for your very informative video. In my work it is not very common to display box and whisker plots to management. It is more common to show long-term average values, monthly average (I run scenarios in a model and do comparative analysis)
Thanks, Nicole! There is definitely a risk in only showing averages and not the "shapes" of distributions, though. For example, different data sets can have very different distributions but the same average. Also, I no longer use box plots at all now, opting for other distribution chart types such as strip plots and distribution heatmaps instead; see nightingaledvs.com/ive-stopped-using-box-plots-should-you/
Thx Nick for sharing your knowledge. What I like about ur teaching is how reasonable your arguments are and inspire me immediately when to apply what I learn from you. Subscribed immediately. Was looking for your Beyond dashboards book, but couldn't find it!
Thanks, Abderrahim. The book has been delayed but will be out next year (2022) and it's been renamed "Practical Dashboards": www.practicalreporting.com/practical-dashboards-book-summary
@@practicalreporting thanks for the heads up. Please keep your videos coming- your content and subjects are way different and solid compared to what is already published in the site.
The sample frequency heatmap in this video was mocked up using Excel's conditional formatting feature, with the median lines added manually. I'm sure more clever people than I could figure out a less hacky way to do it in various other applications, probably using stacked bars.
![Understand Box Plots in Statistics (Box-and-Whisker Plots) - [6-8-23]](http://i.ytimg.com/vi/cklvFT8dR64/mqdefault.jpg)
![Understand Box Plots in Statistics (Box-and-Whisker Plots) - [6-8-23]](/img/tr.png)
great video! straightforward and informative. Loved that you also offered an alternative option
Thanks, @Jashmyn!
It was very helpful and easy to understand, thanks for the hard work
Underrated, pretty good explanantion
very useful, thank you and great explanation/
Super helpful! Thank you
what an explanation. I Appreciate. Thank You.
Glad that you found it to be useful.
Oh my God thank you so much I finally understand it
precise and concise
Bro is LEGEND
really nice, thank you.
Amazing video! Thank you!
this explanation helped me, thank youuuu!!!
This is veery useful ty~
great job! thank you!
Glad you found the video to be useful!
Thank you for this !!! :)
No problem 😊
Your explanations were exceptionally clear.
thanks
Great video 👍
amazing
Nice video, but how do you locate extreme values?
I didn't cover it in the video, but there's a widely used convention for determining what's an outlier and showing them in a box plot. This article describes it: www.real-statistics.com/excel-capabilities/creating-box-plot-outliers-manually/
Please how will you report - 1.113 skewness
Well, skewness in general (not just 1.113) will appear in a box plot as the "whisker" and/or "box" sections at one end of the box and whisker shape being shorter than the box and whisker shapes at the other end.
In a distribution heatmap, skewness appears as colored cells at one extremity of a column of cells being darker than cells at the other extremity.
Kind of hard to explain without visual aids...
Okay
Thanks for the video! just one question🙋 Why does the heatmap use 0%, 30%, and 60% instead of 0%, 25%, 50%, and 75%?
Thanks, Janet. This is a bit of a coin toss, to be honest. In general, scales with intervals based on 1, 2 or 5 (integers that are naturally divisible into 10) make it easiest for people to perceive values in a chart so, yes, intervals of 30% (i.e., based on 3) aren't ideal. Intervals of 25% might be better than 30%, but I tend to avoid intervals based on 2.5 (not an integer that's naturally divisible into 10). The question, then is whether intervals of 5%, 10%, 20% or 50% would work better. Of these, only 20% would "fit" in a small scale like the one in the frequency heatmap. There would be a lot of stops on that small scale, though (0%, 20%, 40%, 60%), so it might be crowded-looking. Like I said, a bit of a coin toss between 30% (less intuitive but cleaner-looking) and 20% (more intuitive but more cluttered-looking).
AHHHHHHHHHHHH
aswome
Thanks!
something abouut u says you r canadian
I am!
@@practicalreporting sorry.....i am watching too much of "how i met your mother"
As someone who's studying to become a psychologist this was very useful thank you!
Mr., I have got a question. In which tool did you design the Frequency Heatmap? It was very stylish and clear to interpret data.
It was actually created in Excel, using conditional formatting (and making the numbers in the cells invisible, see support.microsoft.com/en-au/office/hide-or-display-cell-values-c94b3493-7762-4a53-8461-fb5cd9f05c33 )
@@practicalreportingthis will only work if the value we want to represents are inside the table. What if we wanted to see the frequency of that value and represent that in the heatmap?
in this vid that would be for example 20 people with >120k salary. Our data would have 20 different names with >120k salary. To represent that 20 people we need a table of countif salary>120k and we need to do this for each salary band. Then we do the conditional formatting..
CMIIW. Good vid!!!
@marior6662 To make a distribution heatmap, yes, each cell has to contain (or be associated with) the number (or %) of values that fall within that cell. That value then determines the color of that cell. The example that I showed was created in Excel using conditional formatting, and the numbers in the cells were made invisible using this trick: support.microsoft.com/en-us/office/hide-or-display-cell-values-c94b3493-7762-4a53-8461-fb5cd9f05c33#:~:text=Hide%20cell%20values&text=On%20the%20Home%20tab%2C%20click,Type%20%3B%3B%3B%20(three%20semicolons).
Informative & easy to understand! Thanks for the explanation!
You're welcome!
Most easy and most informative!
Hi Nick, congrats on this exemplary explanation about box plot. Curious to know what´s your opinion on why people don´t find this chart very intuitive? (minute 04:20) .... are there any like sociological or maybe anthropological reasons to explain why people have a hard time understanding a very informative plot? Please if you have any links to refer me, I would greatly appreciate it. Lastly, do you make those charts on the video with Excel or Tableau? thanks, mate, greetings from Perú.
Glad that you found the video to be useful. My comment about many people finding box plots to be unintuitive is based on my experiences explaining them to thousands of workshop participants, and the fact that they require an understanding of the abstract notion of quartiles (which very few people possess in most organizations) in order to be interpreted. I suspect that many people find frequency heatmaps to be more intuitive since we pre-attentively associate higher color intensity with higher quantities (in this case, higher concentrations of values), whereas people have weaker pre-attentive associations for box and whisker shapes. Frequency heatmaps also only require an understanding of bins, which are easier to grasp than the concept of quartiles.
If you have a statistical background, you may find these reasons to be kind of silly (i.e., "quartiles aren't that hard to understand..."), but most people don't have statistical backgrounds ;-)
unfortunate that such a relevant channel has 118 subscribers only given that it has been more than an year.
Thanks, Aditya. This channel is mostly for supplementary information for my training workshop participants, so I don't actively promote it.
Thank U
Hi Nick, thank you so much for your very informative video. In my work it is not very common to display box and whisker plots to management. It is more common to show long-term average values, monthly average (I run scenarios in a model and do comparative analysis)
Thanks, Nicole! There is definitely a risk in only showing averages and not the "shapes" of distributions, though. For example, different data sets can have very different distributions but the same average. Also, I no longer use box plots at all now, opting for other distribution chart types such as strip plots and distribution heatmaps instead; see nightingaledvs.com/ive-stopped-using-box-plots-should-you/
Excellent sir.
Thx Nick for sharing your knowledge. What I like about ur teaching is how reasonable your arguments are and inspire me immediately when to apply what I learn from you.
Subscribed immediately.
Was looking for your Beyond dashboards book, but couldn't find it!
Thanks, Abderrahim. The book has been delayed but will be out next year (2022) and it's been renamed "Practical Dashboards": www.practicalreporting.com/practical-dashboards-book-summary
@@practicalreporting thanks for the heads up. Please keep your videos coming- your content and subjects are way different and solid compared to what is already published in the site.
which software you are using to creat frequency heatmaps
The sample frequency heatmap in this video was mocked up using Excel's conditional formatting feature, with the median lines added manually. I'm sure more clever people than I could figure out a less hacky way to do it in various other applications, probably using stacked bars.