Learning "Graphical reporting/Visualization"- Where to start and progress..............?

[novinosrin]

Learning "Graphical reporting/Visualization"- Where to start and progress..............?

Hi All, I have run into a compelling situation at work where the requirement is for me/us in Credit risk portfolio management to build graphical reports using statistical graphics procedures etc. 

I have long been avoiding to learn this for the reason I have been told Tableau overwhelmingly replaces any other visualization tool. Of course, I/We love SAS and in any given day, our emotions are tied to SAS and would want SAS to be the leader in this space too. Well, C'est la vie!

Can you folks suggest me a book/documentation or place to start that pretty much covers from the very basic to advanced concepts. FYI- I do have a paid subscription to O'Reily Safarionline library, albeit just do not know which would be the material/place to get me started. 

Thank you in advance!

[PaigeMiller · Accepted solution]

Way too much content about SAS Graphics for any single human to read is here:

https://blogs.sas.com/content/graphicallyspeaking/

If I may also be so bold as to suggest reading about good graphical design (which has nothing to do with SAS specifically)

https://www.edwardtufte.com/tufte/books_vdqi

Another excellent read: How to Display Data Badly (not free)

https://www.jstor.org/stable/2683253

And finally, The Best Statistical Graphic Ever Drawn

http://www.openculture.com/2019/07/napoleons-disastrous-invasion-of-russia-explained-in-an-1869-data...

[PaigeMiller]

Also, with regards to Tableau, from my perspective in Credit Risk, it is a very effective tool, and people like it. It seems to be most useful for prepared reports that are issued on a regular basis (monthly, weekly, etc.) It is claimed that anyone (with a license to use Tableau) can go into Tableau and call up their own ad hoc reports as needed, which does indeed sound like a major advantage, but at least in my experience I don't know anyone doing that. What Tableau isn't really that good for is when there has to be some data analysis to lead to a conclusion, and this analysis is usually performed by data analysts (often using SAS or Python) who apply statistical and mathematical methods to the task. Yes, there are some statistical methods built into Tableau, but not a lot and at least when I saw what was available in Tableau, the statistical tools in Tableau were weak. And the data analysts can do the analyses in SAS or Python, produce some prototype graphics in SAS or Python, and then if there's a need to make the results more widely available and have this analysis update each time the data updates, then Tableau is the tool. If the graphics are really needed one-time only, as happens on many projects, I don't really see a benefit to Tableau.

But in any case, if you are getting a typical Credit Risk questions, such as "We put into place program ABCDEF and now the data is starting to come in, are we getting the expected benefits?", or "Losses were noticeably higher last quarter and we'd like to know why", I think these are good questions to be solved by SAS or Python and not by Tableau.

[ballardw]

Not going to claim any expertise in Tableau though have been involved with some use.

My biggest concern is anything that lets people "explore" data often opens the window to, being polite, misinformed analysis. Then someone uses "but your data says ..." .

Especially when people try to get "average rate of <anything>".

Canned choices and controlled selection methods to prevent such seem important to me.

Next are the details of analytics. Some of the data I work with is complex survey design collection and I cringe when random subsets or combinations of data are "analyzed" with other tools not using the full design information.

And in a loosely related comment, just because you can make a graph of topic XXX vs YYY doesn't mean you should. I see some of that in my organization's Covid-19 impact website.

[PaigeMiller]

My biggest concern is anything that lets people "explore" data often opens the window to, being polite, misinformed analysis. Then someone uses "but your data says ..." .

Especially when people try to get "average rate of <anything>".

This is a concern of mine as well, people who are not trained in data analysis can come up with crazy wrong analyses and never even realize it. But we are already there, with people doing their own stuff in Excel or JMP or similar.

[Ksharp]

And  @GraphGuy  website .

http://robslink.com/

or @tc could shed a light to you .

[novinosrin]

Thank you @ Everyone for lending your time in providing me with the much needed guidance. 

One last question or concern- Do you reckon SAS graphics is here to stay so that one can value the time and effort needed to learn or would it be wise to just search for examples, survive with example code for the time being and rather watch out for potential Tableau to walk over everything? 

@Reeza  Hope you are well. Requesting your thoughts too plz

[Reeza]

@novinosrin I am well, just back from a weeks vacation that was definitely needed!

I have some questions first though:)

1. Does your company already have a BI tool in place? If so, which tool?
2. Who will be your primary customer for these visualizations? End clients, management at company?
3. Does it need to reflect live data? 
4. What formats are currently used, are reports generated as Word, PDF, HTML documents? 

[PaigeMiller]

---

@novinosrin wrote:

One last question or concern- Do you reckon SAS graphics is here to stay so that one can value the time and effort needed to learn or would it be wise to just search for examples, survive with example code for the time being and rather watch out for potential Tableau to walk over everything? 

---

In my opinion, there will always be ad hoc analyses and prototype analyses that are best done in SAS (or Python or similar) rather than Tableau (or similar). The ability of a true programming language to do the analyses and present graphical results still far outstrips what can be done in Tableau; and furthermore in my opinion, it makes no sense for data analysts to produce an analysis using one of the bazillion statistical and mathematical methods in SAS or Python, and then be forced to use another tool (Tableau or similar) to do the graphics.

[novinosrin]

@Reeza  and others, Please find my responses in italics

1. Does your company already have a BI tool in place? If so, which tool?  We seem to have more than one BI tool. I am aware SQL server-(SSRS,SSIS and SSAS), Tableau and SAS
2. Who will be your primary customer for these visualizations? End clients, management at company? Since I am part of the Credit risk residential lending(Consumer Mortgage risk), much can be be deemed internal, however "internal" would be mean an array of teams closely working teams us like Loss forecasting/Home equity finance/Residential mortgage/MSR(Mortgage servicing etc). BTW, I am into Portfolio management, i.e dealing solely with Performance and trends of Asset quality.
3. Does it need to reflect live data? Elapsed -Weekly, Monthly and Live Daily(liquidity risk reports, these are 3 sets of reports)
4. What formats are currently used, are reports generated as Word, PDF, HTML documents? ODS to PDF document slides. My boss's boss(1 level above is a fan of SAS ODS Graphics)  

Thank you!

[PaigeMiller]

Are there people at your company who do their own ad hoc graphics in Tableau? Or do they simply use pre-programmed Tableau reports?

Are there people at your company who perform a data analytic role, provide programming to perform analysis/model building and graphics on an ad hoc basis as needed.

Is your company stupid enough to mandate that all graphics must be done in one tool (which is Tableau)? Is your company cheap enough and stupid enough to say: Now that we have Tableau, we can get rid of SAS? And where does Python and R fit in?

[novinosrin]

Sir @PaigeMiller  Thank you for the thoughts-

Are there people at your company who do their own ad hoc graphics in Tableau? Or do they simply use pre-programmed Tableau reports?

Tableau at ours is still a recent trend. From what I have seen so far, there are some teams who are increasing their usage gradually, so I am not sure what would be the usage going forward nor in what capacity they are utilizing tableau at this point.

Are there people at your company who perform a data analytic role, provide programming to perform analysis/model building and graphics on an ad hoc basis as needed.

Nobody does an exclusive set of tasks, it's basically need based. People tend to raise their hand to let the management know their comfort level and it goes like that. However, when it comes to visualization, many or if not most are not quite familiar with either of the tools. For some reason though, Tableau seems to be the hot topic at our office whether or not if that's truly worth being the center of all these discussions.

Is your company stupid enough to mandate that all graphics must be done in one tool (which is Tableau)? Is your company cheap enough and stupid enough to say: Now that we have Tableau, we can get rid of SAS? And where does Python and R fit in?

So far no. I hope not! Perhaps I am unreasonably concerned and over-analyzing things needlessly. When I hear people give too much importance to tableau over and over, I wanted to throw open a request to hear comparison with respect to my personal situation. 

[PaigeMiller]

Just so you know, @novinosrin, I have worked for companies that were stupid enough or cheap enough to make the decisions I described. Fortunately, I am not referring to my current employer.

[Reeza]

IMO, Tableau and tools are best used when you have regular reporting for operational work. Support for decision making is definitely a use case, but I find that decisions often need new data because they reflect new problems. That means constantly rebuilding visualizations which take too long in a full BI implementation (test, migration to prod, scheduling, working with IT).  

My recommendation would be to learn SGPLOT and some of the basic ODS options to control your graphics - image format, location, size etc. 

Use SAS to prototype reports quickly and then when a report is finalized, develop it into a BI solution once it's been running for a while without issue. And hopefully that's someone else's job as well 🙂 `

Since PDF and slides are still your primary usage, that's one area where SAS quality graphics will be better, especially if you need to continuously update a slide deck each month. Tableau doesn't really support that as easily as you can have a custom PPT generated quite easily from SAS. 

If you want to get into graphics, I'll echo the recommendations to learn some best practices as well. In addition to Edward Tufte, I'll recommend Alberto Cairo, Nathan Yau (FlowingData) and Stephen Few.

[PaigeMiller]

---

@Reeza wrote:

IMO, Tableau and tools are best used when you have regular reporting for operational work. Support for decision making is definitely a use case, but I find that decisions often need new data because they reflect new problems. That means constantly rebuilding visualizations which take too long in a full BI implementation (test, migration to prod, scheduling, working with IT).  

My recommendation would be to learn SGPLOT and some of the basic ODS options to control your graphics - image format, location, size etc. 

Use SAS to prototype reports quickly and then when a report is finalized, develop it into a BI solution once it's been running for a while without issue. And hopefully that's someone else's job as well 🙂 `

Since PDF and slides are still your primary usage, that's one area where SAS quality graphics will be better, especially if you need to continuously update a slide deck each month. Tableau doesn't really support that as easily as you can have a custom PPT generated quite easily from SAS. 

If you want to get into graphics, I'll echo the recommendations to learn some best practices as well. In addition to Edward Tufte, I'll recommend Alberto Cairo, Nathan Yau (FlowingData) and Stephen Few.

---

I agree with everything you just said, @Reeza, although I might have chosen a different emoji 😋