SAS Support Communities

"Correlations does not imply causation." Does that bring back memories from  your college statistics class? If you cringe when you hear those words, don't worry. This phrase is still relevant today, but is now more approachable and easier to understand.   At SAS, we use SAS Viual Analyticts to make sense of it. We can use a correlation matrix to explore relationships between variables and forecasting to figure out which variables explain a response or target variable.   Before we take a look at that, let’s first dig into how forecasting works in SAS Visual Analytics. Although the business user may not necessarily know this, SAS Visual Analytics runs both Exponential Smoothing Models (ESM) and Auto Regressive Integrated Moving Average models – ARIMA, for short.    If those sound scary, all you really have to know is that they predict future data as a function of the historical data values. Time series models aren’t the same as simply extending a linear trend. Recent data points are weighed more heavily when calculating the future data points. Makes sense, right?   So we have ESM and ARIMA models in SAS Visual Analytics. For a simple forecast, using a line chart in the SAS Visual Analytics Explorer, and without choosing any underlying factors (independent variables), SAS Visual Analytics calculates the Root Mean Square Error (RMSE) for each ESM model and selects the one with the lowest RMSE from the following:   Damped-trend exponential smoothing Linear exponential smoothing Seasonal exponential smoothing Simple exponential smoothing Winters method (additive) Winters method (multiplicative) ESM Models are effective, but don’t include underlying factors in the forecast. ARIMA models can include these and are called ARIMAX models when they do.   When you select underlying factors, SAS Visual Analytics initially selects one ESM and two ARIMA models. It then calculates the RMSE for each model again and, you guessed it, selects the one with the lowest RMSE as the best model.   After all this magic happens in the background, you’ll notice that some underlying factors are grayed out and some are not. If all are grayed out, it means that the selected model is an ESM or ARIMA model. If there are one or more significant underlying factors, then the selected model is an ARIMAX model.   These significant underlying factors can add to the accuracy of the forecast, and the data points for these factors can be moved up or down using the Scenario Analysis capability.   Now that you have a general idea of how forecasting works in SAS Visual Analytics, let’s see how this relates to correlations. I’ve been working with lots of financial services companies lately, so I’m using some vehicle insurance data in my examples. Here’s a correlation matrix I ran. The tooltip of the tile I’m hovering over shows me that there’s a strong relationship between the Risk Score variable and the Travel Time to work variable.   This makes sense on an intuitive level as well: the more time you spend on the road, the higher your risk score should be from an insurance perspective. However, what's important to note here is that this strong correlations of 0.8540 only describes the strength of the relationship, but tells me nothing about cause and effect.           Enter the forecast with underlying factors. If I add Risk Score as my variable to forecast and drag in most of the measures available in my data set, I only see Claim Count, Gross Losses, and Time in Force (number of years as a customer) as my underlying factors that have an influence on risk score.     Now, keep in mind that these can change depending on adding or deleting the underlying factors. The moral of the story is that we have a clear example that correlation and forecast results do not necessarily have to match because correlation does not imply causation.   Just because my Risk Score and Travel Time to work variables are highly correlated, does not mean that Travel Time to work causes a high risk score. As intuitive as it may seem, the underlying factors are based on statistical significance, not on what makes sense from a business point of view. Understanding, even at just a high level, the inner workings of forecasts helps me reconcile this in my head and feel confident that I’m providing others with accurate results. And to me, that’s very comforting.   If you’re interested in learning more about SAS Visual Analytics or SAS Visual Statistics, what better way to do so then by trying it out for yourself? Don’t forget to let us know what you think!   ... View more

Here at SAS, we often get questions about how to manipulate date and date/time variables in SAS Visual Analytics. Working with dates may seem tricky, but it doesn’t have to be if you keep in mind some simple functions that are available in SAS Visual Analytics.   Below I've summarized the most frequently asked questions we get about dates and provided steps on how to get answers to these. They may seem basic, but can go a long way when you’re working on reports that require date manipulation. For some more advanced calculations using parameters, check out Renato’s post on Date Parameters in SAS Visual Analytics.   1) Is there an easy way to figure out the difference in X between a date and the previous time period?   Yes! SAS Visual Analytics can automatically calculate the difference in a measure between a date and the previous time period, or, between a date and the parallel time period for a larger interval. For example, the difference in measure X between January and February of the same year, or, January of year Y and January of year (Y – 1) can be calculated by doing the following:   Right click on the measure you’re interested in finding the difference for and select Create. Then, simply select the difference you’re looking for based on the date you want to use.       2) How do I calculate the number of days between two time periods?   It’s simple. In the Report Designer, create a new calculated item. Name your calculated item and drag over the x-y function under Numeric (simple). Next, drag over the TreatAs. You can find this under Numeric (advanced). You can use this function when you have a datetime value and want SAS Visual Analytics to treat it as a different data type. In this case, a numeric one, as shown below. Drag over your start and end dates and save the calculated item. You can now use it in report objects such as a list table to easily find out how many days you have between two dates.     To find out the difference between today’s date and another date in your data, drag over the DatePart function, which can be found under Date and Time. Assign the Now function to it, which can also be found under the Date and Time section. Since the output of this is a date, you will have to use TreatAs again to wrap around DatePart (Now). Then, simply subtract TreatAs (your target date) with number as the output. To clean up the calculated item, select Numeric as its format once you’ve created it. Here’s what all of this looks like:       3) How can I allow the report consumer to pick a historical date and have my data filtered accordingly?   You already know that sliders work well for time ranges, but if your report user wants to see historical data for a particular date that falls into a start date and end date range, you can allow them to input this date in a Text Input box. You can take advantage of the Parse function here to filter the dates based on the user’s input.   First, create a new parameter called Choose Date and assign that as the parameter for your Text Input box. Then, drag over the Parse function (under Text (simple)), which will interpret the input string as ANYDTDTE9. if you set it to that format. You can then use BetweenInclusive, which is a Comparison operator to select the start and end dates variables based on your data. Note that for the Parse and BetweenInclusive operators to work together, they both need a date format. Here is what this filter looks like:     Once you have the filter set up, you can now test it by typing in a date in the Text Input box. If you create a list table with a few chosen variables, including the dates, of course, you will be able to see the list table filtered accordingly.       There are many more variations of these calculated items available. These are just the basics to help you get started with date calculations. Remember that TreatAs, Parse, the Comparison operators, and the DatePart operator are all great ones to play around with when you’re working with dates in SAS Visual Analytics.   What other tricks and tips for dates have you used? ... View more

Background: This document describes the integration between SAS Enterprise Guide and SAS Visual Analytics. Note that SAS Enterprise Guide supports integration with SAS Visual Analytics 6.2, 6.3, 6.4, 7.1 and higher. Note on older versions: The first maintenance release for SAS Enterprise Guide 6.1 (which shipped in December 2013) provides integration with SAS Visual Analytics 6.2 and 6.3. Because of this integration, you can open and refresh reports that were created using SAS Visual Analytics Designer and SAS Visual Analytics Explorer. You can also save a SAS Visual Analytics report with your SAS Enterprise Guide project. Note that you cannot generate SAS Visual Analytics reports in SAS Enterprise Guide.   Opening a SAS Visual Analytics Report in SAS Enterprise Guide To open a SAS Visual Analytics Report using the File dialog box, you select File à Open àReport and change the “Files of type” control to be “SAS Report (2G) Files”, as shown in Figure 1 below. To open a report using the SAS Folders panel in the Resources pane, set the Show field to either “All applicable SAS types” or to “Report (2G)”, as shown in Figure 2 below. You must have the data set used to create the report loaded in order to view the report in SAS Enterprise Guide. An administrator can help with this. Figure 1. File dialog box in SAS Enterprise Guide     Figure 2. SAS Folders panel in SAS Enterprise Guide     Upload to LASR task in SAS Enterprise Guide The upload to LASR task is available in SAS Enterprise Guide 7.1 and 7.11. The purpose of this task is to upload your data to the SAS LASR Analytic Server so that you can use it in SAS Visual Analytics. Before you execute this task, you must have the LASR server started. A SAS Visual Analytics administrator should be able to verify that the LASR server is up and running.   Figure 3. The Upload to LASR task in SAS Enterprise Guide   Figure 4 shows the library, table name, and location that can be specified when you use the Upload to LASR task:   Figure 4. Upload to LASR details   A word of warning about this process: ‘Upload to LASR’ doesn’t guarantee that the data file will continue to be in LASR Memory if somebody (such as the SAS Visual Analytics administrator) ends up unloading it or the Server goes down/reboots. A possible more preferred flow is to have SAS Enterprise Guide write directly to the SAS Visual Analytics AUTOLOAD location on the Server.   Export SAS Visual Analytics Report to PDF In SAS Enterprise Guide 7.1, you can export a SAS Visual Analytics report to PDF, as shown in Figure 5:   Figure 5. Export to PDF option in SAS Enterprise Guide     Extract data from SAS Visual Analytics reports in SAS Enterprise Guide Once you have a report open in SAS Enterprise Guide, you have the ability to extract the data. This can be done by visualization, as shown in Figure 6. Figure 7 shows the row and column filter options you have when you extract the data.     Figure 6. Extract Data option in SAS Enterprise Guide   Figure 7. Extract Data filtering options in SAS Enterprise Guide     Integrate Stored Processes from SAS Enterprise Guide into SAS Visual Analytics reports Stored processes are SAS programs that can be executed when and where needed by other applications. SAS Visual Analytics reports can consume stored processes. When creating a report, simply add an existing stored process to the report via the Objects tab in the report designer, as shown in figure 8.   Figure 8. Stored Process option in SAS Visual Analytics Report Designer   The actual stored process can be created in SAS Enterprise Guide, as shown in figure 9 below. To look for a stored process in SAS Enterprise Guide, navigate to the SAS folder pane and then “Show: Stored Process”.   Figure 9. SAS Stored process example in SAS Enterprise Guide. ... View more