From here on in, we assume you have a tidy dataset. That is all the dataset setup has been done. We focus now on the analysis and reporting.
Making a table
If you are in the Variable Summaries view: Select a variable in the Accordion, and then drag it on top of another variable.
- You will notice the cursor changes to the open hand icon when you hover over a variable in the Accordion.
- Left-click and drag the variable over to the right… without letting go (yet).
- Notice how the screen changes in the Main Area with lots of different rectangular slots - these are called drop zones.
- Drag your variable over another variable (which will have become a drop zone) and drop it there.
If you are in the Tables and Graphs mode: Select a variable in the Accordion, and drag it into the Columns drop zone.
You can select other variables to go into the rows simply by clicking on them in the Accordion. Or, you can drag them into the Rows drop zone.
Customizing tables and graphs
At the bottom of the screen is the all-important Display Controller. It has lots of options to help you change the look of your table. It will only appear if you move your mouse to the bottom of the screen.
Let’s start by looking at the chart types. Crunch will only display chart types that are appropriate to the table you are working with (so you can't make a crosstab into a donut chart). Simply click the chart type you want. You can change it at will and even turn it back into a table.
You can also change the statistics in the table.
Most typically, researchers want to change between the Counts (N) and the Percentages (%) - but there’s also the Z-score and Index available on crosstabs.
The next button across is an arrow that tells us which direction percentages are calculated (rows or columns).
The default is the down-arrow, which means the percentages are column percentages.
But we can change that. If we did the horizontal arrow, it would become a row percentage. Most commonly, researchers use the down arrow because the explanatory variables are in the columns. The square box icon is if you want a total percentage (that is, each cell's proportion of the entire valid sample in the table).
Hypothesis testing (tests of statistical significance)
Next, the asterisk * icon turns the default statistical testing on and off.
Where the cells in the table are now shaded according to this legend in the lower-left corner of the screen.
Note as well a scale legend key opens down the bottom left.
We'll explain briefly here what these mean, but you should refer to our introductory article if you want to learn more.
Crunch presents a scale of p-values. The darker the color, the smaller the p-value, that is, the more confident we are. The direction is indicated by color. Green cells are significantly higher for that cell than what we'd expect, and red are lower.
The default shading is based on the Z-score which takes into account the whole table (not just comparing two numbers). It shows how far a cell deviates from expectation. It takes into account the whole table (ie: row and column expectations). This overall shading identifies cells of interest. From there, you can opt to investigate further as per the below.
In the example above, it shows that the 23% for China is significantly higher than if we were to combine all the non-China columns into one. Some people like to think of it as comparing China against the total (although that's not technically correct - again read our introductory article to learn more).
Whereas this default coloring gives a view of the overall difference between rows and columns, sometimes we just want to compare other columns to a specific column. That is equivalent to pairwise column comparisons (like the ABCD letters).
For example, let’s say we want to compare everyone to Germany in the image below.
- Hover the mouse directly under the column
- Notice it says Set Comparison underneath the column
- Click this to set comparisons
Now Crunch is comparing the column in grey to everything else pairwise. So in the example above, it’s comparing the 19% for Germany to the 23% for China. China in this case is statistically higher (at the 0.1 level) than Germany. France is not highlighted compared to Germany because p>0.1 in every case.
You can move the Set Comparisons around to different columns. This is equivalent to exploring ABCD letters like in traditional tab books, but you are focused on one column at a time here. This can feel odd for new researchers, but the advantage here is that you don't have to deal with a long string of letters - it is clear in the interactive environment what is being compared to what. If you do want ABCD letters, you can achieve this later on in your exports to Excel.
You can turn off hypothesis testing altogether by again clicking the asterisk in the Display Controller.
You change the precision of the decimals with the button that looks like 0.00:
The double arrows button allows you to switch the rows/columns for a table or chart.
Finally, the three dots on the far right open a panel that lets you toggle other display preferences. It also lets you access additional table manipulations (such as moving averages for date questions).
Sorting Tables and Graphs
Tables can be sorted by clicking on the column headings (once for descending, again for ascending, and then again to return to the original variable order):
Crunch will show a small chevron to indicate the sort direction. Once you've sorted your table, simply switch to another visualization type to see that sort order reflected in the categories. The analysis will remain sorted in this way, even if you apply filters (i.e., the categories will change order, as applicable, to retain the largest-to-smallest sequence):
Single-dimension analyses (i.e., those that show just one variable) can also be sorted after they've been saved to the deck. In the edit panel for a saved analysis, there is a "Categories" tab and we've added a new toggle option here for switching between Manual, Descending, and Ascending:
Dragging and dropping any category into a new position will automatically revert an analysis to being manually sorted.
The known limitation of this first version is that we’re not offering the ability to ‘anchor’ a category to a position – most commonly used to keep “Other” or “None” etc. categories at the bottom. For now, the only way to achieve this is the previous method of manual dragging-and-dropping. We intend to offer category anchoring in the future.