MARC SCHAFFER: Greetings.Thank you for joining for this video today, where we'regoing to talk about how we can best visualize regressionanalysis and specifically the output of regression analysis.[Marc Schaffer, Program Coordinator, Director, Centerfor Business Economic Analysis]I'm Dr. Marc Schaffer, an AssociateProfessor of Economics at St. Norbert College.I'm also the Director of our Centerfor Business and Economic Analysishere in addition to being the Coordinator for our DataAnalytics Program.I'm really excited to be talking to you today
MARC SCHAFFER [continued]: about a very important topic that I'vegrown to become very passionate about,which is how we can best visualize our research.Early on in my career I spent a lot of timelearning about different methodologies,how we can best analyze and run regressionon all different sorts of things.But we didn't really spend a lot of timetalking about how we can best present this information.
MARC SCHAFFER [continued]: For me, this video is really a topic of,if I knew then what I know now, whatwould I do differently in terms of howI pursued things early in my careerfor presenting my research?That is what I want to talk about today.A quick disclaimer before we get started--this is really a video about how we can bestpresent the results of regression,not a video on how to do regressionor how to analyze or interpret the analysis.
MARC SCHAFFER [continued]: So the assumption is you're somewhatfamiliar with this topic.We'll put some links along with this videothat you can dig a little bit deeper into this topicif you want to.But again, the assumption is goingto be that you're already familiar with whattypical regression output will look like.[Presenting Your Research at Conferences]To help bring the idea to life today, what I want to do
MARC SCHAFFER [continued]: is spend some time presenting some of my own failureswhen it comes to presenting research.So I want to talk about a paper that I'vepresented that I worked on with some colleagues,that we later got published, that we presentedduring graduate school.And I want to give you a context of this paper first,show you some of those results, and then show youhow I don't think this was the best way to goabout presenting our research.So just a quick 30-second clip of what
MARC SCHAFFER [continued]: this paper is about-- myself and some colleagueswere looking at the impact of executive compensation,on the risk-taking behavior of a firm.The idea here being, does it matterhow we compensate executives?If we compensate them with short-term bonuses,for example, they're being rewarded for quick profitstoday, not necessarily looking for the long term, whichmight change their risk behavior of their firm.
MARC SCHAFFER [continued]: If we compensate them with long term incentives like stockoptions, this is rewarding them for longer term stabilityof the firm performance.So we were trying to understand, does itmatter how we compensate those executives?So some of the quick findings thatare going to be relevant for what I'llshow you momentarily-- the biggest thing we foundis that the long-term incentives do actuallyreduce the risk-taking behavior the firm, whichis a good thing.
MARC SCHAFFER [continued]: We found that these results were strongerafter the financial crisis of 2007 and 2008.And lastly, one of our big value adds,which caused a lot of challenges for presentingthis information, was we looked to see whether or notthese effects were consistent across various industries.This created the challenge because thisput a lot of variables in that wehad to interpret and present, which made
MARC SCHAFFER [continued]: it very difficult to do so.Well, I wanted to provide some context to the paper.The other thing I wanted to highlight in this videois the idea that a lot of times research--we think about how things have always been done.We think about some standards in the industry.And earlier in my career, I approached a lot of my researchin the framework of, I would write the paper,
MARC SCHAFFER [continued]: I would put all the information in the paper,I would cross every T, dot every I.And then, when I would go do the presentation,I would look at the paper, and I would pull things outof the paper, and then I would pull the tables or the figuresor the graphs directly from the paper,and I would put them right in the slide deck.It made the process of making a presentation relatively simple.But over time you start to realize that might notbe the best approach because what we want to do in a paper
MARC SCHAFFER [continued]: might be different from how we might want to presentthis visually in that context.We've all been to conferences where we've seen presentations,and the presenter gets up and says,at the beginning of a slide or something,oh, this might be really hard for you guysto see in the back, or I know this is a lot of information,and this is overwhelming.Any time you hear presenters say something like this--and I've said this before myself--what that probably means is maybe
MARC SCHAFFER [continued]: there's a better way to represent thisthan what you're seeing here because Ihave to give you a disclaimer.So I'm going to show you the first table in this paperthat we worked on.And please withhold some judgmentbecause this table's a little bitrough to be honest with you.This table, as you can see right here--it's a terrible table to present in a slide.There is way too much information on this table.It's difficult to read, partially because a lot of it
MARC SCHAFFER [continued]: is way too small.And it's overall very overwhelming for the audienceto take a look at and interpret.In fact, there's so much information on this tablethat you almost need the presenter thereto walk you through each component of the table.That's a bit of a problem.The other problem that comes about is when you're actuallypresenting this type of information, often what happens
MARC SCHAFFER [continued]: is the audience will automaticallyspend a lot of time trying to interpret or readall the information on the slide that they're probablynot listening to you as the presenter.This is basically not great for you getting your pointacross of what your research is trying to tell you.So the real question here is, is therea better way to possibly present this information?
MARC SCHAFFER [continued]: This is one of those things that broader data visualization--we often overlook how important itis in our academic preparation because we don't spendenough time talking about it.How can we leverage data visualizationto better get the meaning of our results across to an audiencein a physical presentation?So that's what we want to talk about is,what are some strategies for better presenting
MARC SCHAFFER [continued]: your statistical analysis and your statistical results?[Four Approaches to Presenting Regression Output]In many ways, academia and research,by definition, we're known to be very innovative.That's what we do as a field.And while that's true in many ways,one of the things I've come to find in this field
MARC SCHAFFER [continued]: is that there's still a lot of, well, thisis the way we've always done this, specificallywhen it comes to presenting our research.Often our papers-- they should look a certain way.Our tables should look a certain way.We should use this font, but we shouldn't use this font.There's a lot of things about howwe present this information that'sbecome somewhat of an industry standard in many ways.Now, this makes a lot of sense whenwe're talking about publishing our work in a journal that
MARC SCHAFFER [continued]: has specific standards of how thingsshould look to abide by what the general look and feel is.But often this doesn't translate very wellto an in-person presentation format.The goal of the in-person presentationis really to convey your key takeaways as clearlyand succinctly as you possibly can in the shortest frame
MARC SCHAFFER [continued]: that you can.And often this requires a good amount of creativitythat's a little different than what I would put in my writtenjournal article because when we write the written journalarticle, what we're trying to do is do a complete in-depthanalysis so anybody that's reading that paper--they know all the background, all the context,all the extra robustness checks that we did.All the information needs to be there because you're not there
MARC SCHAFFER [continued]: to explain it.If you're doing an in-person presentation,you're there to bring the information, the material,the content, the research to life.That's what you're trying to do.So specifically, let's talk about how we currentlypresent a lot of this research.And let's discuss a couple of techniques and tipsthat we can use to enhance the effectivenessof our presentations.So specifically, what we want to dois talk about four approaches to presenting our regression
MARC SCHAFFER [continued]: output.Now, these are just approaches we can start with.They don't all have to be mutually exclusive.There's different ways to do everything we're talking about.But I want to talk about four ways that we can approach this,specifically-- first, the most common thing that we seeis the full table approach.So we'll take a table directly from our researchand stick it right in a slide deck.We'll talk about what that looks likeand why that might be a good idea
MARC SCHAFFER [continued]: and why that might not be a good idea in certain contexts.Then I want to talk about a layering approach,and we present, how can we layer our presentation to better tellthe narrative of our research?Then I want to talk about possiblytruncating your tables to highlight important variables.And lastly, and most importantly for this video,how we can use a visualization approach via coefficient
MARC SCHAFFER [continued]: plotting to better tell the story of our research.[Full Table Approach]So the first approach I want to talk aboutis the full table approach.This is exactly what I showed you earlier in the videofrom the research that I originally was presenting.And this is the idea of taking whatever
MARC SCHAFFER [continued]: the full table is from a research,we can stick it right in the slide.This is a common, classic way to present regression analysisin that standard table format.And often if you have a very small modelwith only a couple of variables, this approachis perfectly fine.However, if you actually have a lot of variables,this can get very overwhelming very quickly,which is problematic when you're trying
MARC SCHAFFER [continued]: to get your key points across.So what I do is want to pull this table again.Let's talk about it for a few secondsand talk about why, in this context,the full table approach might nothave been the best way to go.So if you look at this table, the first thing yougot to realize when looking at regression outputis there's a lot of standard conventions and thingsyou see when you're showing the results of your analysis.So a couple of the key things you'regoing to look for in any regression table
MARC SCHAFFER [continued]: is first, what's the key dependent variable?So if you look at the table that I'm working with here that I'mshowing you, on the top you see the natural logof sigma, which is the natural log of the standard deviations.That's the risk variable that we'retrying to understand in this study that I'm working with.And then on these tables, which we'll commonlysee down a single column, is all the independent variables.So these are all the variables that we'retrying to understand how they impact that dependent variable.
MARC SCHAFFER [continued]: And alongside each one of those independent variables,what you'll commonly see is whatever the coefficientestimate is from the regression outputand then usually some indication aboutwhether or not this is statisticallysignificant in terms of its p value most commonly.Typically you'll see, as in the case of the tableyou're looking at here--typically you'll see asterisk, for example,noting the statistical significance.
MARC SCHAFFER [continued]: That's what we used in this paper.And this is what a typical regression outputtable is going to look like.Now, as I mentioned before, if you're using the full tableapproach, if there is a small number of variables,this is perfectly fine for a slide,and it's easy to interpret, easy to highlight the key takeaways.But as you can see in this table here--partially because we did a lot of work lookingat a bunch of different industries--
MARC SCHAFFER [continued]: we had a lot of variables, so this tablebecomes very overwhelming to the audiencethe moment they look at it.If you look at the second model that we estimated,the right hand side column there,there's about 28 variables in there.So there's a lot of data for the audience to interpret.So in this case, the full table approachdoes not work very well for getting our key points across.
MARC SCHAFFER [continued]: [Layering Approach]So is there a way that we can improve upon this approach?Well, the first thing I often suggest--and this is for presenting tables, but just presentinganything for that matter--is to use a layering approach.This is a really good way to keep the audience focusedon what you want them to look at when you're talking about it.
MARC SCHAFFER [continued]: So the idea of layering is just simply taking a single slideof data that you would normally present as a slideand breaking it up into different componentsor different layers so you can guide the audiencethrough the presentation.So here's one way we can use this idea whenleveraging the full table that I already showed you.So if I was giving this presentation again,and I wanted to use the full table approach, what I would do
MARC SCHAFFER [continued]: is show you the full table like I am here--this is the table you've already seen.And suppose I wanted to have a conversation just the datain the first column.What I could do then is advance to the next slide,as you can see here.And what I'm actually going to do is cover the second columnand you just see the single first column.This allows the audience to focuson just the numbers I want them to be looking at while I'm
MARC SCHAFFER [continued]: talking about them, so they're not actuallyscanning the rest of the table trying to understand.And they're listening to me talk about the specific thingI want them to see.I could, if I'm still using this table,advance to the next slide, and I could hide column one.And I could then highlight column twoand use that to talk about the key thing I want themto see at that moment in time.Now, this is a simple idea of layering
MARC SCHAFFER [continued]: where you're just literally guiding themthrough a particular slide or particular tableset or something like that.This works really well for any size of regression outputtable.The one thing I will say is, while I'mdoing this in this context by hiding different columns,you could also do this by putting arrows or boxesor highlights in some way to draw the audiencefocus to something in particular on the slide.It's not a very complicated thing
MARC SCHAFFER [continued]: to do, but actually makes a huge differencefrom the presentation standpoint of getting the audienceto follow the narrative that you want them to tell.[Truncated Table Approach]So while the layering approach canhelp us get a better idea of how wecan tell our story of our research,
MARC SCHAFFER [continued]: that still doesn't solve the initial problemthat we started this video with which is that the table justhad way too much information, was hard to interpret,and it was very overwhelming.So one thing we can often do when presenting our researchis actually take a truncated table approach,which is the third thing I want to discuss.And this is not a complicated idea.The reality is, oftentimes when we're doing our research,
MARC SCHAFFER [continued]: there's a lot of things about whatever our question isthat we already know.There's already been existing research on it.There's already some key variables we know matter.And we have to include them in our study.But realistically, that's not the valueadd that we're bringing with our research.So one thing we could do is, if weknow we have all these variables that are control variablesthat we know already matter and we're not
MARC SCHAFFER [continued]: overly concerned about presenting themwhen we're talking in the presentation,we want to focus on highlighting just the variables thatare most important that we added value to with our research.So one thing we can do is just simply truncate the table.So I'm pulling up a visual here, whichis a truncated version of the table you've already seen.And what you've noticed is I took a whole chunkof the middle variables out.
MARC SCHAFFER [continued]: And in the first column, you onlysee two independent variables that we looked at in this modelbecause those are the two variables thatare most important for me getting the point acrossabout our research.This is a really simple thing to do, and all you have to dois highlight the key variables.It is important to note that if you use this approach,you also should highlight somewhere on that tablethat you did have control variables that you included
MARC SCHAFFER [continued]: in the model, you're just not showing them.But this is a really good way to helpfocus the audience's attention.Now, while this is a good approach,we still have the same problem we've alreadyhad with this crazy table that I've been working with,which is that if you look at column two,there's still a ton of variables there.We were looking across various industries,so we had a lot of information and a lot of variables.
MARC SCHAFFER [continued]: How can we actually make this betterwhen I'm presenting it in person so it's notoverwhelming with all this information?[Visualizing Output Approach]For our last approach, we want to thinkabout how we can best visualize our regression output.And the idea here is really thinking
MARC SCHAFFER [continued]: about why we visualize data to begin with.As economists we spend a lot of time thinkingabout costs and benefits.And when I think about data visualization,it really is all about efficiency.How can we best visually convey the most meaning that we canin the fastest amount of time, if it's a presentation,or in a way that has the least amount of interpretationor processing on behalf of the audience?
MARC SCHAFFER [continued]: How do we efficiently present this information?This is where we, as researchers,can leverage visualization to better represent our regressionoutput.And specifically, the way that we can visualize our regressionoutput is to use something called a coefficient plot.Now, if we're going to talk about the best wayto possibly visualize regression output,there's a couple of key things wedo have to mention about what we're trying to tell
MARC SCHAFFER [continued]: when we run a regression.Commonly, we're concerned with two particular things.When we're looking at the independent variables,we want to look at the sign, whether or notit has a positive or negative relationshipwith our dependent variable.And we also want to look at whether or notit's statistically significant.That tells us that this effect can actually be interpretedand it's not the case where there'sno impact in this situation.
MARC SCHAFFER [continued]: So if that is the objective, if that'sthe thing we're trying to highlightwith our analysis and our research,a coefficient plot is a great way to be able to do this.Now, essentially what a coefficient plot is going to dois it's going to plot each one of your coefficient estimatesthat you would have seen on the table.But rather than show the p value for statistical significance,what it's going to show is the confidence interval.
MARC SCHAFFER [continued]: Now, this is going to inform the same informationthat you get from the table with the p value,only now you can see it visually.So what I've done here is I've pulled upa visual of the first column of that tablethat we've already looked at several different times.And I'm showing, in this context, a coefficient plot.Now, the nice thing about these isthey're really easy to interpret and visually fast to interpret.
MARC SCHAFFER [continued]: So specifically what you're going to look atis, each one of the dots on this plotrepresents the coefficient estimate, the lines--the confidence interval.And what you see as I've indicated on there wherezero is.If the confidence interval actuallyintersects with the zero line, then weknow that this is not statisticallydifferent from zero, so that result can be interpreted.If we see that that confidence interval does not
MARC SCHAFFER [continued]: intersect with zero, we know that that valuecan be interpreted.So in this case, we were looking at the impactof the short-term compensation by way of bonusesand the impact of long-term incentives like stock options.And we're looking at those two key independent variables.Now in this case, it's really easy to see,when you look at this plot, that those top twovariables-- percent bonus and percent long-term incentives--
MARC SCHAFFER [continued]: it's easy to see that these are both negativebecause the dots are to the left of zero,and they're statistically different from zero.This is very quick to interpret, whichmakes it nice as a way to present this information.Another advantage to using this visual approachof coefficient plots to show your outputis if you have multiple models that you're trying to show.So if you think back to the table that I showed you before,
MARC SCHAFFER [continued]: we had two different models that we estimated.And we had column one and column two,and we had a lot of coefficient estimates,and we had a lot of stars everywhere to showstatistical significance.One thing we can do visually is actually justplot both models on the same plotbut highlight the difference in those coefficient effects usingcolors, for example, to highlight this.So here you can look at a plot where
MARC SCHAFFER [continued]: we're looking at the impact of executive compensationon firm risk.This is the same thing you just saw,but now I've added the second model to the story.So here you see that the column one model that we lookedat in the table is in blue.The column two model is in red.And I'm just looking at the top half of the variablesin this visualization.But it's really easy to see if those results aresimilar across these two models.It's easy to see if they're negative
MARC SCHAFFER [continued]: or if they're positive, if they're statisticallysignificant or they're not.And this makes this really nice to show multiple modelsat the same time, specifically to see if there's differences.So in this context, you can see, if you look down at the tradevariable, for example, this is one where it was positiveand significant in the original model-- you see that in blue--but it was actually not statisticallysignificant in red in that second model.
MARC SCHAFFER [continued]: This is where you can really see the valueof this type of approach to presenting our analysisbecause you can pick up a lot of those differences thatare harder to see if you're lookingat a table full of numbers.The other place where these plots can add a lot of valueis they're really good for breaking upcomplex or heavy tables into smallerparts for visualization.So as you've noticed, one of the big challengeswith the research that I'm presenting here
MARC SCHAFFER [continued]: is because we're looking across industries,we had a lot of these interactive terms where we'retrying to look at how the bonus effect worked in the miningindustry versus the manufacturingindustry versus whatever industry we were looking at.So we had this long list in our tableof all these different possible interaction effects.This is difficult to interpret because therewas just a lot there.So one thing that I could have done
MARC SCHAFFER [continued]: that would've been a lot better way to present thisis to actually break that up into two smaller tablesusing coefficient plots.So if we take a look at this visual I pulled up here,what I've done is taken that bottom half of the second modelfrom our table and I've broken it up into two figures.The first figure on the left hereis going to show the bonus share, whichis looking at how each one of the different industries
MARC SCHAFFER [continued]: have a different effect between short-term compensationand the risk variable.And I'm also showing you the same informationfor the long-term incentives.Now, for the way that this model was set up--just for context of what you're looking at--all these results are interpreted relativeto the financial services industry.But if you know that, when you look at the first figure,the bonus year figure, what we quickly see
MARC SCHAFFER [continued]: is we know that mining and transportation, communication,the electricity industry-- those actuallyhave a more positive effect of bonus share on firm riskrelative to the financial services industrybecause those two are the ones that are different.It's quick to see that, and you don'thave to scan a table of 6, 7, 8, 9, 10 numbersto get that interpretation.If we look at the other figure here, the long-term incentive
MARC SCHAFFER [continued]: share one, again it's quick to seethat there's four industries here that are actuallymore positive in this effect relative to the financialservices industry by just skimming this table once.It's easy to see.This is where we see the value of coefficient plots.And these work really, really wellfor the in-person presentation.[Visualizing Regression Output, Tips Tricks]
MARC SCHAFFER [continued]: Now that we have a basic understandingof these coefficient plots and a visual waythat we can represent our regression output,if I've convinced you that this is somethingyou want to think about doing for your presentation, whatare some tips and tricks or what aresome things you should think about if this is somethingyou're planning on pursuing?The first thing I will tell you is, above all, when presentingany kind of regression analysis, keep it as simple as possible
MARC SCHAFFER [continued]: to convey your meaning.If you'd only have two or three independent variablesthat you're looking at, a full tableis often perfectly fine because there'snot a ton of information to overwhelm.You have to ask yourself if it's worth the timeto build a data visualization if it's notgoing to add a lot of value based on a couple of numbersthat you already have in a small table.The second thing I'll say is always,
MARC SCHAFFER [continued]: and this is for any presentation for that matter, use layeringwhenever possible in conjunction with whatever approach you'reusing to present your data because it'sgoing to better tell the story of your research.If you're going to use coefficient plots,the other thing I often tell people to think aboutis look at the scales of your coefficients.Sometimes the way that your variables are set up,you might get coefficients that have wide-ranging values.
MARC SCHAFFER [continued]: If that's the case, coefficient plots don't necessarilywork that great because the bottomaxis will be all over the place, and some numbersmight be really small, some numbers will be really big.And it might be difficult to interpret.If you have this issue, and you dowant to do coefficient plotting, look at waysto possibly standardize your dataor standardize your coefficients in a way thatactually makes it easy to interpretwith a coefficient plot.
MARC SCHAFFER [continued]: The fourth thing I'll suggest is,if you're doing anything where there's multiple models you'rerepresenting on a single plot, alwaysthink about using various pre-attentive attributes.Maybe use different colors or different shapes,different ways to highlight or indicatethat there's multiple things you'reseeing in this visualization, different models, for example.The last thing I'll tell you is to think about beingas efficient as possible.
MARC SCHAFFER [continued]: Again, if you only have one or two variables,maybe the table is the way to go.But if you're going to make these,I wouldn't worry about trying to create themfrom scratch manually.Oftentimes, several of our statistical programsthat actually use coding and various functions likethat, a lot of them actually have pre-cannedfeatures where they can automatically createcoefficient plots for you.So if you're going to do this, I'd
MARC SCHAFFER [continued]: recommend you spend some time figuring outhow to actually build those coefficient plots.Because oftentimes, it's an additional linein a command after a regression analysisthat can actually show you how to do this.[Tools Resources]So as I mentioned in the last point,there's a lot of tools and ways that we can actually
MARC SCHAFFER [continued]: make coefficient plots if that's somethingwe want to do for visualizing our regression output.Most visualization tools are able to build these graphsand it's not too difficult to do.Specifically, statistical coding languages--often if you're coding in commands, so if you use Stata,for example, or if you use R, for example,you have your command that runs whatever itis that you're trying to run.
MARC SCHAFFER [continued]: So if you're doing R, it's the LM() function will run a linearmodel to do your regression.There's actually a package that builds offof that called coef_plot.It stands for coefficient plot in R. And all you have to dois use this command and you can call the modelthat you just estimated.And it will automatically build you one of these plotsthat you can then tweak and formatas you need to make the output look however
MARC SCHAFFER [continued]: you want to make it look.The same thing holds in a program like Stata.There's a coefplot function.And after you run your regression,you can essentially call this with coefplot,and it can spit out a pretty tablelike the ones I've showed you here.If you spend a little bit of timeto figure out how the program you use does this,it can pay huge dividends over the courseof your career for presentations because you build itinto something you would normally do.
MARC SCHAFFER [continued]: You estimate your model, you fit the model,you do all the standard tests that youdo to make sure that you're happy with the model you'veestimated.And literally it's a matter of just adding a single commandto spit out these nice, pretty visualizations that you canuse in your presentations and possibly in papersas well depending on what you're trying to do.One more quick note on the layering concept--there's a lot of tools that can do this.Whatever presentation software you're using,
MARC SCHAFFER [continued]: you can always build in layers.A lot of this is how you want to do it.If you work in PowerPoint for example, it's really easy to--in the case of what I showed you here,if you're going to present a table of information,you can actually just drop shapes over that tableto blank out parts of the table youwant to show and just make it different slides in a row.And you can easily layer in your presentations.
MARC SCHAFFER [continued]: And this is just as applicable in something like Beamer.It's really just thinking creative about howyou built your slides.And often when you do use layering--I will say I give a lot of public presentationson state of the economy and I use a lot of layersbecause there's a lot of data and depth to the analysis.One of the things that I've noticed over timeis your slide deck will likely be extremely large because you
MARC SCHAFFER [continued]: might take what was a single slideand break it up into three or four or five piecesto get layers in to better tell your narrative.But I will say it is often well, well worth the timeand effort to layer because your audience is very much goingto appreciate you walking them through the narrative of whatyou want them to know.[Conclusion]
MARC SCHAFFER [continued]: Now that we've spent a good amount of time talkingabout how we can visualize our regression output,what are some of the key takeaways of this video?The first thing I want to stress is, giving presentationsis a norm in academia.It's a norm in our field as researchers, as academics.This is one of the things that we do.So something that is super important is to think about,how do we best convey our results and our research
MARC SCHAFFER [continued]: in a way that people can follow and understandso we can get our key points across?Oftentimes, the easiest way with our statistical analysisis to use tables.That's something we're used to doing.That's a norm in a lot of our fields.However, depending on the size of these tablesand what your analysis looks like,this might not be the best approach.So I encourage you to think about more creative ways
MARC SCHAFFER [continued]: that you can better present your research.At the end of the day, you spent a lot of time and efforttrying to do some kind of analysisthat you are very passionate about.Don't sell it short by not taking the timeto figure out a good way to present itso people can follow what you're trying to tell them.I've talked about four different approachesin this video of how you can present regression output.
MARC SCHAFFER [continued]: Some of them are standard-- so obviously use the full table,using a truncated table to highlightthe important variables, and even usinglayering to better tell the narrative of your research.But the important takeaway here for this videois, how can you use data visualization to better bringyour research to life?The example we use talked about the coefficient plot, which
MARC SCHAFFER [continued]: is a great way to do this.And this visualization can have a meaningful impacton your audience and is especiallygood at comparing coefficients across different modelsand breaking up really data-heavy tablesinto nice, digestible chunks.There's a lot of standard ways that wecan do this and standard packageswe can use to build these kinds of plots.I would encourage you, especially
MARC SCHAFFER [continued]: if you're in the early stage of your career,to spend some time figuring out howto do this because it will definitely pay dividendsthroughout the remainder of your career.Thanks for joining me.And I hope you enjoy making some coefficient plotsin the future.
MARC SCHAFFER [continued]: nbsp;
Publisher: SAGE Publications, Ltd.
Publication Year: 2021
Keywords: charts (data visualization); communication aids; data visualisation; graphical presentation of data; regression; regression coefficient; research findings; Scatterplot; Statistical packages; tabular data ... Show More
Segment Num.: 1
Marc Schaffer, Director at Center for Business & Economic Analysis, and Associate Professor of Economics at St. Norbert College, discusses the coefficient plot to visualize regression data including presenting research at conferences, four approaches to visualizing regression data, tips, tricks, tools, and resources.
Looks like you do not have access to this content.
Marc Schaffer, Director at Center for Business & Economic Analysis, and Associate Professor of Economics at St. Norbert College, discusses the coefficient plot to visualize regression data including presenting research at conferences, four approaches to visualizing regression data, tips, tricks, tools, and resources.