Skip to content

Click in-app to access the full platform documentation for your version of DataRobot.

Filter data

A Data Filtergram™ is both a filter that helps you transform your data and a histogram that helps you visualize your data. Filtergrams allow you to visualize your data before, during, and after every transformation. You can use Filtergrams to:

  • Explore your data. When you use filters in this way, you can view your filtered selections in the Preview Display. Your selections are not saved to steps in the Project. Rather, they let you quickly explore your data. When you're done with your exploration, you can remove the filters and continue with transformation operations in your Project.

  • Remove rows. If your data prep work involves removing rows, you need to first use filters to identify the rows you want to remove.

  • Change the data in just a particular set of rows. In this case, you use filters to isolate the rows where you want to make changes. Then you can apply your changes to just those rows. For example, if you have a column with first names and you want to change all instances of "Anna" to "Ann," then you filter on the column to display only the "Anna" values. Then you apply a Change into Custom Value column operation to convert "Anna" to "Ann."

  • Publish just a particular set of rows. In this case, use filters to isolate the rows you want to publish. Then you can add a lens to create the publishing point.

Create a Data Filtergram

To create a Data Filtergram:

  1. Locate the column you want to filter.

  2. Hover over the column menu icon and click Filter values.

Note that you can have Filtergrams for multiple columns open so that you can preview the results of those filters on the Preview Display.

There are five types of Filtergrams that correspond to each type of column data:

The following sections describe how to use filters of each type.

Text Filtergrams

The Text Filtergram pane displays a list of each distinct text value that appears in your dataset. The bars running left-to-right provide a histogram view of the relative number of times each value occurs. The total number of unique values is listed on the bottom left side of the pane. From the list, you can select values to dynamically display in your dataset.

See Work with Text Filtergrams for actions you can take with Text Filtergrams.

a. Currently Selected (upper left): When you make selections from the list, the button’s label changes to indicate your number of selections. Click the button to open a new pane that lists every selected value. From this pane, you can continue to refine the text values you want to filter on in your dataset. The actions you can take from this pane are described in the section Work with Text Filtergrams.

b. Sort list order: by default, the list of text values displays from highest to lowest count. To reverse the order to display least to most, click the triangle in the upper right corner above the count column. You can also sort the list alphabetically by clicking the triangle located above the list. The triangle’s orange color indicates which sort order (numeric or alpha) is currently applied to the dataset.

c. Color-coded filter bar: When moving your mouse anywhere over the Filtergram pane, the following buttons display:

  • Type: When you mouse over this button, a ratio displays to express the count of currently selected text type rows relative to all text type rows in the dataset. If nothing is selected in the list, the ratio reflects the total count for text type rows relative to all rows in the dataset. When you click this button, you dynamically hide these text values in your dataset. This is useful if you have blanks, errors or other non-text values in this column and want to view only those data types.

  • Other: If there are non-text values in the column, for example numbers, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with non-text values relative to all rows with non-text values in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with non-text values relative to all rows in the dataset. Click Other to toggle off and hide the other values from the current dataset view.

  • Blank: If there are blanks in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected blank rows relative to all blank rows in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of blank rows relative to all rows in the dataset. Click Blank to toggle off and hide the blank cells from the current dataset view.

  • Error: If there are errors in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with cell errors relative to all rows with errors in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with cell errors relative to all rows in the dataset. Click Error to toggle off and hide the cell errors from the current dataset view.

If you have "other" values, blanks or errors in the column, notice there is also a + button adjacent to the button for that value. When you click the + button you add all values of that type, for example all "other" values, to the "Selected Items" list where you can continue working with them. See Work with Text Filtergrams for the filtering actions you can take from Selected Items.

Work with Text Filtergrams

The Text Filtergram pane provides tools that enable you to dynamically filter your data with great precision. This section describes the actions you can take.

a. Select values from the list to display in your dataset: Click any text value in the pane to dynamically filter your dataset to display only that value. To select more than one value: CTRL+click (Windows) or Command+click (Mac). To select a contiguous range: Shift+click. To remove any selection: Alt+click.

b. Show Selected Items: After selecting text values for filtering, click Show Selected Items. A new pane opens and displays every selected Range and Value currently displayed in your dataset. Note: a Range for text fields is defined by ASCII sort order. From this pane, you can take the following actions:

  • b1. Manually remove any of the Ranges or Values in the dataset by clicking to remove the orange check mark adjacent to the Range or Value. Click to return that range or value to your filtered dataset. Note: a trash can icon appears at the top of the pane when you deselect a range or value. This allows you to discard the selection from your filters.

  • b2. Manually update any of the Ranges or Values listed here by clicking the value you want to edit. The value then becomes editable. Enter a new value and click Save.

  • b3. Manually add other Ranges or Values for filtering the data by clicking the + button. Provide a Min and Max for the Range, or provide a Value, and click Save. Your dataset dynamically updates to reflect your additions.

  • b4. Click EXCLUDE to exclude a Range or Value from the dataset. This is particularly useful if you have already selected a range. From the range, you can exclude specific values (from within that range) to hide from the current dataset. Note that ranges and values marked with EXCLUDE are displayed with orange, dotted outlines in the histogram to remind you of their exclusion.

  • b5. When you are finished working in this pane, click Hide Selected Items to return to the list view.

c. Clear and Invert: Clear removes all of your current filters. Invert displays all of your data except that which you have selected to filter.

You can also search for values. To do so, click the magnifying glass icon in the upper right to open a search field. Search for a value that contains, equals, or starts with any value you specify:

Numeric Filtergrams

When you open the Numeric Filtergram pane, there are two views that you can use for performing filtering operations on your data:

  • Show as Graph (the default view) This histogram displays a numeric histogram representing the distribution of numeric values in your dataset. The horizontal line, or x-axis, represents the range of values that occur in your dataset within this column. The height of each bar represents the count of each distinct value in the column. The scale along the left side, or y-axis, provides a measure for the count. The filtering actions you can take with the histogram are described below in the section Work with Numeric Filtergrams.

  • Show as List Click this tab to hide the numeric histogram and display the unique occurrence of every value in the column. From the list, you select values to dynamically display in your dataset. The filtering actions you can take with the list are described in the section Work with Numeric Filtergrams.

When moving your mouse anywhere over the Filtergram pane, the following buttons display:

a. Currently Selected button (top left corner): When you make selections on the histogram, the button’s label changes to indicate your number of selections. Click the button to open a new pane that lists every selected Range and Value currently selected in the histogram. From this pane you can continue to refine the numeric values you want to filter on in your dataset. If you already know the ranges and values that you want to filter on in your dataset, you can click this button instead of using the histogram. From the new pane you can enter values and ranges to begin your filtering operations. The actions you can take from this pane are described in the section Work with Numeric Filtergrams.

b. Log button (bottom left corner): Toggles on a logarithmic scale (log) view of your data. By default, a linear view of the data is displayed in the Filtergram. However, if you have a large numerical range in which one or a few points in the data are much larger than the bulk of the data, the log view adjusts for skewness in your data.

c. Show Zoom/Pan (bottom left corner): Toggles on and off an overview tool used for viewing the relative position of values and ranges that you have zoomed into on the numeric histogram. Zooming and panning operations are described below in the section Show numeric columns as graphs.

d. When moving your mouse anywhere over the Filtergram pane, the following buttons display:

  • Type: When you mouse over this button, a ratio displays to express the count of currently selected numeric type rows relative to all numeric type rows in the dataset. If nothing is selected on the histogram, the ratio reflects the total count for numeric type rows relative to all rows in the dataset. When you click this button, you dynamically hide these numeric values in your dataset. This is useful if you have blanks, errors or other non-numeric values in this column and want to view only those data types.

  • Other: If there are non-numeric values in the column, for example text values, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with non-numeric values relative to all rows with non-numeric values in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with non-numeric values relative to all rows in the dataset. Click Other to toggle off and hide the other values from the current dataset view.

  • Blank: If there are blanks in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected Blank rows relative to all Blank rows in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of Blank rows relative to all rows in the dataset. Click this button to toggle off and hide the blank cells from the current dataset view.

  • Error: If there are errors in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with cell errors relative to all rows with errors in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with cell errors relative to all rows in the dataset. Click this button to toggle off and hide the cell errors from the current dataset view.

If you have "other" values, blanks or errors in the column, notice there is also a + button adjacent to the button for that value. When you click the + button you add all values of that type, for example all "other" values, to the "Selected Items" list where you can continue working with them. See Show numeric columns as graphs for the filtering actions you can take from Selected Items.

Work with Numeric Filtergrams

The Numeric Filtergram provides tools that enable you to dynamically filter your data with great precision. This section describes the actions you can take.

Show numeric columns as graphs

  • Select ranges to view on the histogram: click and drag your mouse across a range of values, and notice that your dataset updates to reflect your selection. To select additional, non-contiguous ranges in the histogram, use the following key commands while dragging your mouse: CTRL+click (Windows) or Command+click (Mac). To remove any selection or portion of a selection, use the following key command while dragging your mouse: Alt+click.

    You can also drag your mouse over the tips of ranges in the histogram to display only those values in your dataset. The y-axis is helpful in determining the relative values of peaks in your data:

  • Begin exploring and transforming your data:

a. Pan across the histogram: Move your mouse over the values on the x-axis. Notice your cursor changes from the pointer to the click and drag symbol. Click and drag across the values on the x-axis to pan across the histogram and your selections. Click Recenter to return the histogram to its default view. Note that your selections are retained.

b. Zoom into a selected range: Move your mouse over the values on the x-axis and use the mouse wheel to zoom into the selection. For a Mac, drag two fingers down to zoom into your selection; drag two fingers up to zoom out. You can continue selecting ranges while zoomed in. To return the histogram to its default view, click Recenter. Note that your selections are retained after the histogram is re-centered.

c. View the relative position of a zoomed range: After zooming into a selected range, you can simultaneously view where that range is located relative to the entire histogram. Click Zoom/Pan to open a second, overview tool histogram below. The overview tool displays the entire range of data and highlights your zoomed selection with a red box. From the overview tool histogram, you can also:

  • Drag the red box to view, at the same zoom range, other values that you have not selected.

  • Work interactively with both the overview tool and the top, primary histogram. Remember that actions you take on the overview tool histogram are only reflected on the primary histogram. To dynamically filter your data, you must make your selections on the top, primary histogram.

d. Show Selected Items: After filtering for selected data ranges, click Show Selected Items. A new pane opens where you specify the exact ranges and values you want to view in your dataset. From this pane, you can take the following actions:

  • d1.) Manually remove any of the Ranges or Values in the dataset by clicking to remove the orange check mark adjacent to the Range or Value. Click to return that range or value to your filtered dataset. Note: A trash can icon appears at the top of the pane when you deselect a range or value. This allows you to discard the selection from your filters.

  • d2.) Manually update any of the Ranges or Values listed here by clicking the value you want to edit. The value then becomes editable. Enter a new value and click Save.

  • d3.) Manually add Ranges or Values for filtering the data by clicking the + button. Provide a Min and Max for the Range, or provide a Value, and click Save. An entry is created for the Value or Range. Click the check box for an entry and your dataset dynamically updates to reflect your selection(s).

  • d4.) EXCLUDE a Range or Value from the dataset. This is particularly useful if you have already selected a range. From the range, you can EXCLUDE specific values (from within that range) to hide from the current dataset. For example, select range 1-2000 to view in your dataset. Then EXCLUDE value 195. The dataset displays everything between 1 and 2000, except for value 195. Note that ranges and values marked with EXCLUDE are displayed with orange, dotted outlines in the histogram to remind you of their exclusion.

  • d5.) Click Hide selections to toggle off the pane and return to the histogram where your selections are highlighted in the histogram.

e. Invert your selection(s): Displays all of your data except that which you have selected to filter.

f. Clear: Removes all of your current filters.

g. Search for values: Click the magnifying glass icon in the upper right corner to open a search field. You can search for a value that contains, equals, or starts with any value you specify.

Show numeric columns as lists

The List view allows you to see the count of every numeric value in the column. Viewing your data in this way is particularly useful when you want to quickly select and filter on specific numeric values. Note that selections you make in the List are reflected in the histogram when you click the Show as Graph tab. This section describes the actions you can take.

a. Sort list order: By default the list of values displays from lowest to highest count. To reverse the order to display highest to lowest, click the triangle in the upper right corner above the count column. You can also sort the list numerically by clicking the triangle located above the list. The triangle’s orange color indicates which sort order (occurrence or numeric) is currently applied to the dataset.

b. Select values to dynamically display in your dataset: Click to select values from the List. To select more than one value: CTRL+click (Windows) or Command+click (Mac). To select a contiguous, multi-row range:Shift+click. To remove any selection, use the following key command: Alt+click. After making selections, click Currently Selected in the top left corner to open a new pane where you can continue refining your filtering operations. If you already know the exact ranges and values you want to use for filtering your dataset, you can skip making selections from the List. Instead, click No Items Selected to open a new pane where you enter exact values and ranges.

c. Refine your selections: After making selections, click Show Selected Items in the top left corner to open a new pane where you can continue refining your filtering operations. The button's label updates to reflect the number of items you have currently selected.

If you already know the exact ranges and values you want to use for filtering your dataset, you can skip making selections from the List. In this case the button's label is No Items Selected. Click the button to open a new pane where you enter exact values and ranges. The filtering operations you can perform from the pane are described in steps d1-d5 in Show numeric columns as graphs. You can also Invert and Clear your selections, and perform a search for specific values from the List. See steps e-g of the same section for details.

Date and Time Filtergrams

When you open the Date/Time Filtergram pane, there are two views that you can use for performing filtering operations on your data:

  • Show as Graph (the default view)

    This histogram displays the distribution of date/time values in your dataset. The horizontal line, or x-axis, represents the range of date values that occur in your dataset within this column. The height of each bar represents the count of each distinct date value in the column. The scale along the left side, or y-axis, provides a measure for the count. The filtering actions you can take with the histogram are described below in the section Show date and time columns as graphs.

  • Show as List Click this tab to hide the date/time histogram and display the unique occurrence of every date/time value in the column. From the List, you select values to dynamically display in your dataset. The filtering actions you can take with the List are described below in the section Show date and time columns as lists.

Work with Date and Time Filtergrams

When moving your mouse anywhere over the Filtergram pane, the following buttons display:

a. Currently Selected (top left corner): When you make selections on the histogram, the button’s label changes to indicate your number of selections. Click the button to open a new pane that lists every selected Range and Value currently selected in the histogram. From this pane you can continue to refine the date/time values you want to filter on in your dataset. If you already know the ranges and values that you want to filter on in your dataset, you can click this button instead of using the histogram. From the new pane you can enter dates and ranges of dates to begin your filtering operations. The actions you can take from this pane are described below in the sections Show date and time columns as graphs and Show date and time columns as lists.

b. Log (bottom left corner): Toggles on a logarithmic scale (log) view of your data. By default, a linear view of the data is displayed in the Filtergram. However, if you have a large date range in which one or a few points in the data are much larger than the bulk of the data, the log view adjusts for skewness in your data.

c. Show Zoom/Pan (bottom left corner): Toggles on and off an overview tool used for viewing the relative position of values and ranges that you have zoomed into on the histogram. Zooming and panning operations are described in the section Show date and time columns as graphs.

d. When moving your mouse anywhere over the Filtergram pane, the following buttons display:

  • Type: When you mouse over this button, a ratio displays to express the count of currently selected date/time type rows relative to all date/time type rows in the dataset. If nothing is selected on the histogram, the ratio reflects the total count for date/time type rows relative to all rows in the dataset. When you click this button, you dynamically hide the date/time values in your dataset. This is useful if you have blanks, errors or other non-date/time values in this column and want to view only those data types.

  • Other: If there are non-date/time values in the column, for example numbers, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with non-date/time values relative to all rows with non-date/time values in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with non-date/time values relative to all rows in the dataset. Click Other to toggle off and hide the other values from the current dataset view.

  • Blank: If there are blanks in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected blank rows relative to all blank rows in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of blank rows relative to all rows in the dataset. Click Blank to toggle off and hide the blank cells from the current dataset view.

  • Error: If there are errors in the column, this button displays. When you mouse over this button, a ratio displays to express the count of currently selected rows with cell errors relative to all rows with errors in the dataset. If nothing is selected on the histogram, the ratio reflects the total count of rows with cell errors relative to all rows in the dataset. Click Error to toggle off and hide the cell errors from the current dataset view.

If you have "other" values, blanks or errors in the column, notice there is also a + button adjacent to the button for that value. When you click the + button you add all values of that type, for example all "other" values, to the "Selected Items" list where you can continue working with them. See below for the filtering actions you can take from Selected Items.

The Date/Time histogram provides a number of powerful tools that enable you to dynamically filter your data with great precision. This section describes the actions you can take.

Show date and time columns as graphs

  • Select ranges to view on the histogram: Click and drag your mouse across a range of values, and notice that your dataset updates to reflect your selection. To select additional, non-contiguous ranges in the histogram, use the following key commands while dragging your mouse: CTRL+click (Windows) or Command+click (Mac). To remove any selection or portion of a selection, use the following key command while dragging your mouse: Alt+click.

    You can also drag your mouse over just the tips of ranges in the histogram to display only those values in your dataset. The y-axis is helpful in determining the relative values of peaks in your data:

  • Begin exploring and transforming your data:

a. Pan across the histogram: Move your mouse over the values on the x-axis. Notice your cursor changes from the pointer to the click and drag symbol. Click and drag across the values on the x-axis to pan across the histogram and your selections. Click Recenter button to return the histogram to its default view. Note that your selections are retained.

b. Zoom into a selected range: Move your mouse over the values on the x-axis and use the mouse wheel to zoom into the selection. For a Mac, drag two fingers down to zoom into your selection; drag two fingers up to zoom out. You can continue selecting ranges while zoomed in. To return the histogram to its default view, click Recenter. Note that your selections are retained after the histogram is re-centered.

c. View the relative position of a zoomed range: After zooming into a selected range, you can simultaneously view where that range is located relative to the entire histogram. Click the Zoom/Pan button to open a second, overview tool histogram below. The overview tool displays the entire range of data and highlights your zoomed selection with a red box. From the overview tool histogram you can also:

  • Drag the red box to view, at the same zoom range, other values that you have not selected.
  • Work interactively with both the overview tool and the top, primary histogram. Remember that actions you take on the overview tool histogram are only reflected on the primary histogram. To dynamically filter your data, you must make your selections on the top, primary histogram.

d. Show Selected Items: After filtering for selected ranges, Show Selected Items. A new pane opens and displays every selected Range and Value currently displayed in your dataset. From this pane, you can EXCLUDE a Range or Value from the dataset. This is particularly useful if you have already selected a range. From the range, you can EXCLUDE specific values (from within that range) to hide from the current dataset. For example, you have selected the following range of dates to display in your dataset: 03/01/2015 - 03/15/2015. You then EXCLUDE the date 03/10/2015. The dataset displays everything in the range except for 03/10/2015.

Note that ranges and values marked with EXCLUDE are displayed with orange, dotted outlines in the histogram to remind you of their exclusion.

A trash can icon appears at the top of the pane when you deselect a range or value. This allows you to discard the selection from your filters.

Click Hide selections to toggle off the pane and return to the histogram where your selections and exclusions are highlighted in the histogram.

e. Available charts: Click this tab to choose from four additional filters—Month of year, Day of month, Day of week, Hour of day—that you can use to filter your date/time data with more precision. After selecting a filter, click the "5 available charts" tab again to pin that filter to your view. While working with any of these filters, notice that your dataset is dynamically updated to reflect your selections.

f. Clear: Removes all of your current filters.

g. Invert your selection(s): Displays all of your data except that which you have selected to filter.

h. Search for values: Click the magnifying glass icon in the upper right corner to open a search field. You can search for a value that contains, equals, or starts with any value you specify.

Show date and time columns as lists

The List filter view allows you to see the count of every date/time value in the column. Viewing your data in this way is particularly useful when you want to quickly select and filter on specific dates. Note that selections you make in the List are reflected in the histogram when you click the Show as Graph tab. This section describes the actions you can take.

a. Sort list order: By default, the list of dates displays from lowest to highest occurrence. To reverse the order to display highest to lowest, click the triangle in the upper right corner above the count column. You can also sort the list chronologically by clicking the triangle located above the date/time values. The triangle’s orange color indicates which sort order (occurrence or chronological) is currently applied to the dataset.

b Select values to dynamically display in your dataset: Click to select values from the List. To select more than one value: CTRL+click (Windows) or Command+click (Mac). To select a contiguous, multi-row range: Shift+click. To remove any selection, use the following key command: Alt+click. After making selections, click Currently Selected in the top left corner to open a new pane where you can continue refining your filtering operations. If you already know the exact ranges and dates you want to use for filtering your dataset, you can skip making selections from the List. Instead, click No Items Selected to open a new pane where you enter exact values and ranges.

c. Refine your selections: After making selections, click Show Selected Items in the top left corner to open a new pane where you can continue refining your filtering operations. The button's label updates to reflect the number of items you have currently selected.

The filtering operations you can perform from the pane are described in step d of the Show date and time columns as graphs section. You can also Invert and Clear your selections, and perform a search for specific values from the List. See steps f-g in the same section.

Boolean Filtergrams

The Boolean Filtergram shows the count of Boolean values in your dataset and filters out any other values from it.

The bars running left-to-right provide a histogram view of the relative number of times each value occurs. The total number of unique values is listed in the bottom left side of the pane. From the list, you can select values to dynamically display in your dataset.

Following is an overview of the Boolean Filtergram display:

Work with Boolean Filtergrams

When moving your mouse anywhere over the Filtergram pane, the following buttons display:

  • Type: When you mouse over this button, the count of Boolean values in this column is listed. When you click this button, you dynamically hide the Boolean values in the column.

  • Other: If there are non-Boolean values in the column, this button displays. When you mouse over this button, the count of non-Boolean values in this column is listed. When you click this button, you dynamically hide the non-Boolean values in the column. Alternatively, click the + button to add all "other" values to the Filtergram list. You can then filter out and hide specific "other" values from the current dataset view by using the following key commands while clicking the "other" value you want to hide: Alt+CRTL (Windows) or Alt+Command (Mac).

  • Blank: If there are blanks in the column, this button displays. When you mouse over this button, the count of blank values in this column is listed. When you click this button, you dynamically hide the blank values in the column.

  • Error: If there are errors in the column, this button displays. When you mouse over this button, the count of cell errors in this column is listed. When you click this button, you dynamically hide the cell errors in the column.

Use Clear and Invert to manage your selections in the Filtergram list. Note that Invert displays all of the values in the current dataset view except the values you have selected in the list.

Source Filtergrams

The Sources Filtergram allows shows how the rows from lookup and appended datasets are participating with the base dataset in your Project.

Source Filtergram example

You have an outer join in which there are rows from a lookup dataset that do not match the base dataset. The Sources Filtergram allows you to see how many rows from the base and lookup datasets are participating in the join. Additionally, a count for the unmatched rows from each dataset source is displayed.

Following is an overview of the Source Filtergram display:

By default, the list of Sources displays from highest to lowest count. To reverse the order to display least to most, click the orange triangle in the upper right corner above the count column. You can also sort the list alphabetically by clicking the triangle located above the list. The triangle’s orange color indicates which sort order (numeric or alpha) is currently applied to the list.

Work with Source Filtergrams

a. Select Sources to display in your dataset: Click any Source to dynamically filter your dataset to display only that Source. To select more than one Source: CTRL+click (Windows) or Command+click (Mac). To select a contiguous, multi-row range: Shift+click.

b. Clear: Remove all Source filter selections.

c. Invert your selections: Displays all Sources except those you have selected.

d. Search for Source files: Click the magnifying glass icon in upper right corner to open a search field. You can search for a Source file that contains, equals, or starts with any text value you specify.

e. Refine your filter selections: Click Show Selected Items. A new pane opens and displays the currently selected Sources. From this pane, you can take the following action:

  • EXCLUDE a Source. This is a toggle that hides the associated Source from your dataset until you click EXCLUDE again. Note that Sources marked with EXCLUDE are displayed with an orange dash in the initial Filtergram pane to remind you of their exclusion.

  • A trash can icon appears at the top of the pane when you deselect a Source. This allows you to discard the Source from your filters.

  • When you are finished working in this pane, click Hide Selected Items to return to the initial Filtergram view.

Dynamic Ranges

The Dynamic Percentiles feature in Filtergrams provides you with the powerful option to specify percentiles of your selected values. For example, if you have an inventory dataset that has a column for products sold per week per region, you can use Dynamic Percentiles to filter select the top 5% highest selling products for each region. Your percentile selections are also dynamically applied to newer versions of datasets that are automatically updated in the Library through Automatic Project Flows (APF). For example, you can use the APF feature with Dynamic Percentiles to automatically produce an AnswerSet each week to identify the top 5% highest selling products for each region per week.

The Dynamic Filtering options are applicable for Date/Time, String, and Numeric type columns. To open the Dynamic filtering options pane:

(a.) Click Currently Selected in the top left corner of your Filtergram.

(b.) Click Add Dynamic Types/Ranges.

The Dynamic Ranges pane opens:

Work with Dynamic Selections

1. Select or deselect the types of values you want to include in your percentile ranges: Valid, Invalid, Blank, Error:

  • Valid: Values that are of the same type as the column type—for example numbers in a numeric type column.
  • Invalid: Values that are not of the same type as the column type—for example alpha characters in a numeric type column.
  • Blank: Use if there are blanks in the column.
  • Error: Use if there are errors in the column.

2. Click plus to add as many different dynamic ranges as desired.

Note

Each of these ranges will be applied to only the types of values that you select in step 1.

3. Set your Dynamic Range: Drag the limits on the number line to set your desired value.

Tip

Use your left and right keyboard arrow keys to adjust for exact values on the number line.

4. (Optional) Add more range selections (following steps 1-3 above). When you create multiple range selections, each additional selection is treated as and AND operation. For example, if you have created two range selections—one set to select the top 10th percentile, and the other to select the bottom 10th percentile—then the values that correspond to these percentiles are highlighted in the filtergram and displayed accordingly on the data grid.

Note

You can always set your dynamic filtering options to work in conjunction with the other filtering operations described in this document for each column type.

5. Order by Value or by Count: when ordered by Value, the percentile is based on the actual values in the column. When you order your data by Count, the percentile is based on the frequency of occurrences per value. For example, you have the following dataset with participant ages and scores on an exam.

If you want to determine how the participants scored in relation to one other, order by Value::

If you want to determine the validity and usefulness of the exam itself or show patterns in the exam results, order by Count::

6. Click Currently Selected to view your highlighted percentiles in the graph or list view.

Behavior of multiple Filtergrams across columns

Filtergrams for multiple columns may be open and dynamically filtered at the same time. Note that the resolution for multiple filtergrams is from left to right—whichever dynamic ranges are set on the left filtergram affect the resulting values that can be selected by the adjacent filtergram on the right:

If you exit out of a Filtergram on the left, then the dynamic ranges being applied for that column are removed.

Note

  • If the Interactive Mode feature is enabled, then your selections are applied to the entire dataset.
  • Dynamic Ranges is a feature that must be enabled. If you do not see this button in your Project, contact your Data Prep System Administrator.

  • Updated October 28, 2021
    Back to top