Results

From QuB

Prev: Histogram	Outline	Next: Scripts

The Results window displays the output from an Action such as Modeling:Idealize or Modeling:MIL. Results include:

"summary" numbers pertaining to the entire computation, such as LL
numbers pertaining to each data selection (or "segment")
any models involved, typically initial and final
histograms

The Results window also has tools for visualization and analysis. You can

see scatter plots, for example segment number vs "occupancy 2" (P_open)
divide segments into color-groups, using the scatter plots or inequalities. Each color-group can be made into a selection list for separate analysis.
use non-linear curve fitting on any scatter plot or histogram
generate a "report" document summarizing the result

1 Life-cycle
2 Pop up window
3 Summary
4 Segments
5 Models
6 Histograms
7 Select
8 Criteria
9 Reports
- 9.1 Options

Life-cycle

Several Actions, such as such as Modeling:Idealize or Modeling:MIL, generate a Result when they finish. The Result appears in the Results window.

Data files keep track of the last several Results in the Results list. You can select an item in that list to show it in the Results window. Old results are discarded as new ones come in; if you'd like to keep a result indefinitely, right-click its entry and choose "Hold". Data files store their Results in the Session file, which is saved when you close the data file or "Save data".

There is a vestigial "Files" column in the Results window, similar to the Data and Model windows. You should make it as small as possible and ignore it.

Pop up window

You can have multiple views on the same Result. Clicking "Pop up window" makes a new window showing this Result.

it's the most convenient way to see it full-screen (maximize the popped-up window)
you can see two (or more) different histograms onscreen together

Summary

The Summary tab contains numbers that describe the computation as a whole, for example LL, and vectors such as "Gradient". You can select cells to copy and paste into another program.

Segments

The Segments tab contains numbers that describe each data selection (if the data source was "list") or segment (if the data source was "file"). You can select cells to copy and paste into another program. Some common columns:

index: the segment number, starting with 0
bounds: the range of data points
start: start time, in milliseconds
length: duration, in milliseconds

Models

The Models tab lists the rate constants of every model involved in the computation. There's usually an "Initial" model that was used to start the computation, and a "Final" model with optimized rate constants. The results of Modeling:Model Search also include numbers such as LL in the model table.

To sort the table, click a column header. To see a model in the Model window, click its entry.

Histograms

Most Results include histograms; either one per segment or one for the whole file. The bars of the histogram are from the data, while the fit curves are PDFs predicted by the model. This gives you a way to tell, at a glance, whether the computation succeeded.

Use the scroll bar to see all the histograms.

When you point the mouse at a histogram, it shows a tool-tip with the mouse coordinates. MIL histograms' tool-tip also includes time constants.

To zoom in, drag a rectangle (marquee) from top-left to bottom-right of the area you want to see. To zoom back out, drag a rectangle from bottom-right to top-left.

Right-click to get this menu:

Extract: save the histogram as a tab-separated text file
Copy Image
Print Image
Curve Fitting
Components Visible: The PDF curve in MIL histograms is the sum of several exponential components. This option displays all the components as thin gray lines
Extract to Excel: makes an Excel worksheet and chart. MS Office 2000 or newer required

Select

Data heterogeneity is common in single-channel recordings and usually poses problems, especially when kinetic modeling is the ultimate goal. The Select tab allows the user to identify and group different types of segments by coloring them.

The Select tab shows scatter plots and histograms of the data from the #Segments table. Segments are shown as colored dots. You can color the dots to group segments; these color-groups can be made into selection lists for further analysis.

For an example using Select, see Single Channel Kinetics#Heterogeneous Bursts and the related tutorial.

Anatomy

"Show var..." menu, at top left: choose a plot to display; choose any of the columns in the "Segments" table; choose "all" to display all columns; choose "none" to hide all plots; choose "Custom" to plot one column vs. another
"Select" menu, at the top (looks like a button): algorithms to work with colors, described below
"SDs" slider, at top right: changes the number of standard deviations shown on the plots
color palette, along the left side: clicking a point changes it to the chosen color; black is by convention "un-selected"; right-click a color in the palette for a menu (described below)
plots: filling the rest of the window

Plots

When you choose a variable from the "Show var" menu, it's displayed with segment number on the x axis and the variable on the y axis. Each dot represents a segment. All the dots start out red. The thick red line is the mean value of the red segments. The thin red lines are +/- N standard deviations, where N is the number on the "SDs" slider, by default 1. Pointing with the mouse shows a tool-tip with the mean and standard deviation for each color.

To change a segment dot's color, choose the color from the palette then click the dot. You can also drag a marquee rectangle over several dots to color them all.

To see a segment's data in the Data window, right-click it.

Right-clicking in the background of a plot brings up this menu:

By index: the default display mode, with segment number on the x axis
By start time: with segment start time on the x axis
As histogram: segment dots are stacked in their bins; the number of bins is controlled by a slider (to the left of the SDs slider)
Log x axis: (custom plots only); plots log(x-value)
Log y axis: plots log(y-value)
Curve Fitting: for scatter plots and histograms
Copy image: copies the exact onscreen image to the clipboard
Extract to Excel: makes an Excel worksheet and chart. MS Office 2000 or newer required

Select Menu

Clicking the "Select" button brings up a menu for manipulating colors. This menu can be extended with Python Scripts.

All: changes all segments to the color chosen in the palette
None: changes all segments to black
K-Means: uses the mean, sd to re-color all segments.; each segment gets the color of the closest distribution.; only variables shown in plots are used; (see below)
K-Means x5: equivalent to choosing "K-Means" five times; hopefully it will stabilize
Trim: removes outliers; for each segment, if it is outside the thin SD lines of its own color in any visible plot, colors it black
Make selection lists: prompts you to export each color-group as a selection list.; you can export just one color-group from the palette's right-click menu

Palette right-click menu

This menu appears when you right-click a color in the palette. You can add to it with Python Scripts.

Trim: for each segment of this color, if it's outside of the thin SD lines in any visible plot, make it black
Select all: give this color to all segments
Select all between lines: give this color to all segments that are between the thin SD lines in any visible plot
Select black between lines: give this color to black segments that are between the thin SD lines in any visible plot
Extract segments: creates a new Result with only the segments of this color
Remove segments: creates a new Result with only the segments of other colors
Make selection list: creates a selection list in the Data window with only the segments of this color

Using k-means and trim

K-means can quickly classify all segments by example. To use it, start with all segments black (Select -> None). Color one segment from each group with a different color, then Select -> K Means. Each segment is given the color of that example point which matches it best on the visible graphs (charts). Repeat K Means until the colors don’t change, then Trim to discard outliers.

Criteria

Use the Criteria tab to change the Select color of segments. Enter one or more "criteria" such as

segments that are black
segments with "LL per event" "<" "3.0"

then pick a color and click "Change Color". Those segments which match ALL of the criteria will get the new color.

If a variable is plotted in select with "log y axis", Criteria will use log(variable) for comparisons.

Reports

A Report is a document that summarizes the Result. A report is made from a template which is a list of Options, for example "File table; Model image; Model table". There are a couple of built-in templates, and you can save your own. The Options are written as Python Scripts, so it's possible to write your own options. All the built-in options write to a MS Word document, with embedded Excel sheets and charts (MS Office 2000 or newer).

You can save custom report templates; they're stored in the QuB program folder under Reports\.

Clicking the Reports button brings up this menu:

New report...

presents a two-column view with all the Options on the right, and your new template on the left.

Double-click or use the arrows to move options between columns and re-order them.
Edit the template's name then press "Save" to add it to the Reports menu
- templates are saved in Program Files\QuB\Reports\
Click "Run" to generate the report

(...): all built-in and saved templates; choose one to run it
Edit reports...: presents the list of templates so you can edit or delete one

Options

(not all options are available or sensible for a given Result)

Header

writes the algorithm name and the date/time it was run

FileTable

lists all Data files involved, with the following as relevant:

dead time
number of ion channels
number of events
LL
containing folder path

LL Summary: writes the LL, LL per event, and number of iterations
Time Constants: (Tau, Amp) of all the components in the duration histogram's PDF
Select: Stats: Mean and standard deviation of each color, for all variables that are showing in #Select
Search Params: writes the standard options for likelihood optimizers
Initial Model Image: adds a picture of the initial Model
Initial Model Table: writes a table of the initial model's rate constants
Initial Model Calc Table: writes an embedded excel table of the rate constants along with the calculated $rate = C * k_0 * e^{Q * k_1}$ and any relevant experimental conditions. Changing the experimental conditions re-calculates the rate. kEq is the ratio of the forward and backward k0.
Initial Model Constraints: writes any kinetic constraints
Final Model ________: same options as for initial model
Histograms: embeds all histograms as excel charts. Double-click one to get at the data series. Writes time constants below each histogram.
Large histograms: same as Histograms but bigger
Select: Charts: adds a picture of each plot showing in the #Select tab
Segment LL: writes a table of each segment's LL, Gradient, and event count
Segment Rates: writes a table of each segment's LL, rate constants and error limits; (when optimizing segments separately)
Segment Time Constants: writes a table of each segment's time constants; (when optimizing segments separately)
Data Image: adds a picture of whatever is in the low-res (top) panel of the Data window
Segments (all): writes a copy of the #Segments table
Segments (visible): writes only those columns from the #Segments table that are showing in #Select
Color Means (all): writes a table of the mean and standard deviation of all variables, for each color
Color Means (visible): writes a table of the mean and standard deviation of the variables showing in #Select, for each color

Prev: Histogram	Outline	Next: Scripts

Retrieved from "http://www.qub.buffalo.edu/wiki/index.php/Results"