Data table
The data table is the matrix that every analysis reads. It lives on the Data stage of an experiment (the third of the four stages). Each row is one well of the plate, each data column is one colour channel of one photo.
You build the table once, by generating it from processed photos or by importing an Excel or CSV file. After that you can edit cells, apply adjustment coefficients, export it and run analyses on any part of it.
Data Analysis
Review and edit your data, then perform statistical analysis and machine learning.
Analysis Tools
Experiment Data Table(9 rows × 7 columns)
| # | SAMPLE NAME Sample Info A | CELL COORDINATE Sample Info B | Origin Classification C | plate_01.jpg R D | plate_01.jpg G E | plate_01.jpg B F | plate_02.jpg intensity G |
|---|---|---|---|---|---|---|---|
| 1 | Tea A | A1 | China | 182.00 | 121.00 | 64.00 | 118.67 |
| 2 | Tea A | A2 | China | 181.00 | 120.00 | 64.00 | 117.33 |
| 3 | Tea A | A3 | China | 183.00 | 122.00 | 65.00 | 120.00 |
| 4 | Tea B | B1 | India | 151.00 | 98.00 | 47.00 | 96.33 |
| 5 | Tea B | B2 | India | 150.00 | 97.00 | 46.00 | 95.67 |
| 6 | Tea B | B3 | India | 153.00 | 100.00 | 48.00 | 97.33 |
| 7 | Tea C | C1 | China | 171.00 | 113.00 | 59.00 | 110.67 |
| 8 | Tea C | C2 | China | 170.00 | 111.00 | 58.00 | 109.00 |
| 9 | Tea C | C3 | China | 173.00 | 114.00 | 59.00 | 111.33 |
The stage has two parts:
- Analysis Tools (left): dataset overview and one button per method. See the analysis dialog. ADJ opens the adjustments dialog.
- Experiment Data Table (right): the table itself. The header shows the
number of rows and columns and three buttons:
- Import — replace the table with data from a file (Import from Excel or CSV);
- Export — download the table and the analysis results as an Excel workbook (Export);
- Regenerate — open the generation dialog again (Generate data from images).
On a narrow screen the tools panel is shown above the table.
Until the experiment has data, the stage shows two buttons instead of the table:
Data Analysis
Review and edit your data, then perform statistical analysis and machine learning.
No Data Available
You can either generate data from your images or import data from an Excel/CSV file.
Rows and columns
- One row = one well that is assigned to a sample in the plate schema. Empty wells (not assigned to any sample) are not in the table. In the demo above the plate has 3 × 4 wells, 9 of them belong to samples, so the table has 9 rows.
- Row order depends on the generation mode: Group by Sample (rows of one sample together, samples sorted by name) or Grid Order (A1, A2, …, B1, …). Imported tables keep the row order of the file.
- The # column shows row numbers starting from 1. They are the row
numbers of the range notation (
D2:G9). - Every column has a letter in the top-right corner of its header (A, B, C, …, Z, AA, AB, …). Letters and row numbers together give Excel-style references to cells, used in the data range of an analysis.
- The line under the table shows totals:
rows × columns | data channels | unique samples. Unique samples is the number of distinctsample_namevalues, that is the number of samples, not wells.
The dataset overview in the tools panel counts the same things with other words: cells is the number of rows (wells), channels is the number of all columns, including the sample and class columns, and samples is the number of distinct sample names.
Column types
Columns always appear in this order: sample information, classification columns, data columns. The type decides what a column can be used for.
| Type | Header (title / subtitle) | Content | Used as |
|---|---|---|---|
| Sample info | SAMPLE NAME / Sample Info | Sample name from the plate schema | Groups for Group K-Fold CV and replicate averaging; can also be a label column |
| Sample info | CELL COORDINATE / Sample Info | Well position on the plate (A1, B3…) | Reference only |
| Classification | class type (e.g. Origin) / Classification | Class of the sample for this class type | Label column of classification methods |
| Data | photo file name / channel (R, G, B, intensity…) | Colour value of the well on this photo | Features (the data range) |
Details:
- Classification columns come from the class types of the plate schema, one column per class type. Wells whose sample has no class of that type stay empty; supervised analyses skip such rows.
- Data columns are grouped by photo, in the order of the photos in the experiment, and inside a photo by channel: R, G, B, intensity, rg_intensity, rb_intensity, gb_intensity. What a channel contains is explained on Generate data from images.
- Data values are shown with two decimals. The stored value is not rounded.
- Internally a class column is called
class_<type>and a data column<file name>_<channel>(for exampleplate_01.jpg_R). You see these names in the exported Excel file.
Editing cells
Click a cell to edit it. Data cells accept numbers, sample and class cells accept any text.
- Enter or a click elsewhere saves the value, Esc cancels.
- A data cell with text that is not a number gets the value
0. - Editing a sample name or a class changes how rows are grouped and labelled. Use it to fix typos or to relabel a sample.
Experiment Data Table(9 rows × 7 columns)
| # | SAMPLE NAME Sample Info A | CELL COORDINATE Sample Info B | Origin Classification C | plate_01.jpg R D | plate_01.jpg G E | plate_01.jpg B F | plate_02.jpg intensity G |
|---|---|---|---|---|---|---|---|
| 1 | Tea A | A1 | China | 182.00 | 121.00 | 64.00 | 118.67 |
| 2 | Tea A | A2 | China | 181.00 | 120.00 | 64.00 | 117.33 |
| 3 | Tea A | A3 | China | 183.00 | 122.00 | 65.00 | 120.00 |
| 4 | Tea B | B1 | India | 151.00 | 98.00 | 47.00 | 96.33 |
| 5 | Tea B | B2 | India | 150.00 | 97.00 | 46.00 | 95.67 |
| 6 | Tea B | B3 | India | 153.00 | 100.00 | 48.00 | 97.33 |
| 7 | Tea C | C1 | China | 171.00 | 113.00 | 59.00 | 110.67 |
| 8 | Tea C | C2 | China | 170.00 | 111.00 | 58.00 | 109.00 |
| 9 | Tea C | C3 | China | 173.00 | 114.00 | 59.00 | 111.33 |
The table is saved to the server when you click Save Changes at the top of the experiment page, when you switch to another stage, and when you click any method button in Analysis Tools (the table is saved before the analysis dialog opens). Analyses always use the saved table. If you close the page without saving, the edits are lost.
Context menu
Select cells, then right-click inside the selection:
| Command | Shortcut | What it does |
|---|---|---|
| Copy | Ctrl+C | Copies the selection as tab-separated text (Excel and other spreadsheets can paste it) |
| Cut | Ctrl+X | Copies the selection, then clears it |
| Paste | Ctrl+V | Writes copied values starting at the top-left cell of the selection |
| Delete | Del | Clears the selected cells |
On macOS use Cmd instead of Ctrl.
- Paste in the menu is available after you copy or cut cells in the table. Ctrl+V also pastes text copied from another program, for example a block of cells copied in Excel.
- Pasted rows go down from the first selected row and may go below the selection. Pasted columns are limited to the width of the selection: to paste a block of three columns, select at least three columns first.
- In data columns a pasted value is read as a number. Use a dot as the
decimal separator:
12,5is read as12. Text that is not a number becomes0, an empty value leaves the cell empty. - Esc clears the selection.
A cleared data cell is empty, and analyses read it as 0. A cleared class
cell is an empty text, which the analysis treats as one more class (it is
not skipped like a class that was never assigned). To leave rows out of an
analysis, exclude their classes in the
analysis dialog instead of clearing
cells.
Selecting a data range
Drag the mouse over cells to select a rectangle. The bar above the table
shows its reference (Selected: D2:F4), and Clear Selection removes it.
The selection is used by the context menu and the keyboard shortcuts.
The data range of an analysis is also written in this notation: the
feature columns and rows the method uses, for example D1:G9. By default
the analysis dialog takes all data
columns and all rows. To pick a range with the mouse:
- In the analysis dialog, click the cursor button next to Data Range. The dialog hides and the table switches to range selection mode: an info bar appears, sample and class columns are greyed out, and the selection turns green.
- Drag over the data cells you need. Only data columns are selected, even if the drag crosses other columns.
- When you release the mouse, the dialog comes back with the range in Data Range. Cancel in the info bar returns to the dialog without changes.
Experiment Data Table(9 rows × 7 columns)
| # | SAMPLE NAME Sample Info A | CELL COORDINATE Sample Info B | Origin Classification C | plate_01.jpg R D | plate_01.jpg G E | plate_01.jpg B F | plate_02.jpg intensity G |
|---|---|---|---|---|---|---|---|
| 1 | Tea A | A1 | China | 182.00 | 121.00 | 64.00 | 118.67 |
| 2 | Tea A | A2 | China | 181.00 | 120.00 | 64.00 | 117.33 |
| 3 | Tea A | A3 | China | 183.00 | 122.00 | 65.00 | 120.00 |
| 4 | Tea B | B1 | India | 151.00 | 98.00 | 47.00 | 96.33 |
| 5 | Tea B | B2 | India | 150.00 | 97.00 | 46.00 | 95.67 |
| 6 | Tea B | B3 | India | 153.00 | 100.00 | 48.00 | 97.33 |
| 7 | Tea C | C1 | China | 171.00 | 113.00 | 59.00 | 110.67 |
| 8 | Tea C | C2 | China | 170.00 | 111.00 | 58.00 | 109.00 |
| 9 | Tea C | C3 | China | 173.00 | 114.00 | 59.00 | 111.33 |
Each drag gives one rectangle. To combine several blocks, type them into
Data Range separated by commas, for example D1:F9, H1:H9.
The rows of the range decide which wells the analysis uses; the label column is chosen separately in the dialog, so the range does not need to include it. See Analysis dialog for the label column, class filters and column filters.