Merge pull request #229 from ajdapretnar/line-chart-doc

Updated documentation

doc/widgets/correlogram.md

@@ -9,12 +9,17 @@ Visualize variables' auto-correlation.
 
 In this widget, you can visualize the autocorrelation coefficients for the selected time series.
 
-![](images/correlogram-stamped.png)
+![](images/Correlogram.png)
 
 1. Select the series to calculate autocorrelation for.
-2. See the autocorrelation coefficients.
-3. Choose to calculate the coefficients using partial autocorrelation function (PACF) instead.
-4. Choose to plot the 95% significance interval (dotted horizontal line). Coefficients that are outside of this interval might be significant.
+2. Choose to calculate the coefficients using partial autocorrelation function (PACF) instead. Choose to plot the 95% significance interval (dotted horizontal line). Coefficients that are outside of this interval might be significant.
+
+Example
+-------
+
+Here is a simple example on how to use the Periodogram widget. We have passed the [Yahoo Finance](yahoo_finance.md) data to the widget and plotted the autocorrelation of Amazon stocks for the past 6 years.
+
+![](images/Correlogram-Example.png)
 
 #### See also
 

doc/widgets/difference.md

@@ -5,17 +5,23 @@ Make the time series stationary by replacing it with 1st or 2nd order discrete d
 
 **Inputs**
 
-- Time series: Time series as output by [As Timeseries](as_timeseries.md) widget.
+- Time series: a dataset, often as output by [As Timeseries](as_timeseries.md) widget.
 
 **Outputs**
 
-- Time series: Differences of input time series.
+- Time series: differences of input time series.
 
-![](images/difference-stamped.png)
+![](images/Difference.png)
 
-1. Order of differencing. Can be 1 or 2.
-2. The shift before differencing. Value of 1 equals to discrete differencing. You can use higher values to compute the difference between now and this many steps ahead.
-3. Invert the differencing direction.
-4. Select the series to difference.
+1. Select the series to difference.
+2. Order of differencing. Options are first or second order difference, change quotient and percentage change.
+3. The shift before differencing. Value of 1 equals discrete differencing. You can use higher values to compute the difference between now and this many steps ahead. *Invert differencing direction* compute differences from latest to oldest. 
 
-To integrate the differences back into the original series (e.g. the forecasts), use the [Moving Transform](moving_transform.md) widget.
+To integrate the differences back into the original series (e.g. the forecasts), use the [Moving Transform](moving_transform_w.md) widget.
+
+Example
+-------
+
+In this example, we are using the [Yahoo Finance](yahoo_finance.md) data for Amazon stocks for the past 6 years. We pass the data to **Difference** to compute the daily change in the high value of stocks. We observe the change in the [Line Chart](line_chart.md) widget.
+
+![](images/Difference-Example.png)

doc/widgets/line_chart.md

@@ -5,23 +5,24 @@ Visualize time series' sequence and progression in the most basic time series vi
 
 **Inputs**
 
-- Time series: Time series as output by [As Timeseries](as_timeseries.md) widget.
-- Forecast: Time series forecast as output by one of the models (like [VAR](var.md) or [ARIMA](arima.md)).
+- Time series: a dataset, often as output by [As Timeseries](as_timeseries.md) widget.
+- Features: list of attributes
+- Forecast: time series forecast as output by one of the models (like [VAR](var.md) or [ARIMA](arima.md)).
 
-You can visualize the time series in this widget.
+You can visualize the time series in this widget. Note that the *line* option will display the data as a connected line. In case of non-existent dates, the widget will show them, then the line will be drawn over to connect the last known value with the next one. In case of missing values, the widget will not draw the line in the given place, making the chart disconnected. To better see the missing values, we recommend using the *column* option.
 
-![](images/line-chart-stamped.png)
+![](images/LineChart.png)
 
-1. Stack a new line chart below the current charts.
-2. Remove the associated stacked chart.
-3. Type of chart to draw. Options are: *line*, *step line*, *column*, *area*, *spline*.
-4. Switch between linear and logarithmic *y* axis.
-5. Select the time series to preview (select multiple series using the *Ctrl* key).
-6. See the selected series in this area.
+1. Add a new line chart below the current charts. The widget can display up to 5 parallel charts.
+2. Set the type of chart. Options are: *line*, *step line*, *column*, *area*. X button removes the associated chart.
+3. Switch between linear and logarithmic *y* axis.
+4. Select the time series to display (select multiple series using the *Ctrl*/*Cmd* key). Filter enables searching for the desired variable from the list by its name.
 
 Example
 -------
 
-Attach the model's forecast to the *Forecast* input signal to preview it. The forecast is drawn with a dotted line and the confidence intervals as an ranged area.
+The example uses [Yahoo Finance](yahoo_finance.md) data for the previous year. We can observe the data in the **Line Chart**.
 
-![](images/line-chart-ex1.png)
+To see the forecast, we have used the [VAR Model](var.md) for training the model. Then, we have attached the model's forecast to the *Forecast* input signal of the Line Chart. The forecast is drawn with a dotted line and the confidence intervals as an ranged area.
+
+![](images/LineChart-Example.png)

doc/widgets/periodogram_w.md

@@ -5,17 +5,23 @@ Visualize time series' cycles, seasonality, periodicity, and most significant pe
 
 **Inputs**
 
-- Time series: Time series as output by [As Timeseries](as_timeseries.md) widget.
+- Time series: Time series from the File or as output by [As Timeseries](as_timeseries.md) widget.
 
 In this widget, you can visualize the most significant periods of the time series.
 
-![](images/periodogram-stamped.png)
+![](images/Periodogram.png)
 
 1. Select the series to calculate the periodogram for.
-2. See the periods and their respective relative power spectral density estimates.
 
 Periodogram for non-equispaced series is calculated using Lomb-Scargle method.
 
+Example
+-------
+
+Here is a simple example on how to use the Periodogram widget. We have passed the [Yahoo Finance](yahoo_finance.md) data to the widget and plotted the periodicity of Amazon stocks for the past 6 years.
+
+![](images/Periodogram-Example.png)
+
 #### See also
 
 [Correlogram](correlogram.md)

doc/widgets/spiralogram.md

@@ -1,26 +1,38 @@
 Spiralogram
 ===========
 
-Visualize variables' auto-correlation.
+Visualize time series' periodicity in a spiral heatmap.
 
 **Inputs**
 
-- Time series: Time series as output by [As Timeseries](as_timeseries.md) widget.
+- Time series: Time series from File or as output by [As Timeseries](as_timeseries.md) widget.
 
-Visualize time series' periodicity in a spiral heatmap.
+**Outputs**
+
+- Selected Data: instances selected from the plot
+- Statistics: data table with statistics as shown in the plot
+
+Spiralogram is intended for visualizing time series and for comparing attribute values by categorical variables. Instances can be selected from the plot and sent downstream.
 
-![](images/spiralogram-stamped.png)
+![](images/Spiralogram.png)
 
-1. Unit of the vertical axis. Options are: years, months, or days (as present in the series), months of year, days of week, days of month, days of year, weeks of year, weeks of month, hours of day, minutes of hour.
-2. Unit of the radial axis (options are the same as for (1)).
-3. Aggregation function. The series is aggregated on intervals selected in (1) and (2).
-4. Select the series to include.
+1. Units on the circumference. Options are: month of year (12 units), day of year (365 units), day of month (~30 units), day of week (Mon-Sun, 7 units), hour of day (24 units) and all the variables from the data.
+2. Unit of the vertical axis. *Hide inner labels* removes labels on the vertical axis.
+3. Color of each spiralogram section. Default is *Show instance count*. If an attribute from the data is selected, aggregation methods become available. The options are: mean value, sum, product, minimum, maximum, span, median, mode, standard deviation, variance, harmonic mean, geometric mean, non-zero count, and defined count.
 
 Example
 -------
 
-The image above shows traffic for select French highways. We see a strong seasonal pattern (high summer) and somewhat of an anomaly on July 1992. In this month, there was [an important trucker strike](https://www.google.com/search?q=french+trucker+strike+1992) in protest of the new road laws.
+In the example below we are using *Illegal waste dumpsite in Slovenia* data, available from the Datasets widget. The data record when the dumpsite was registered, what kind of waste was deposited, whether the site was cleaned, and so on.
+
+We will use [As Timeseries](as_timeseries.md) for set *Entry creation date* as our time series variable. Then we pass the data to **Spiralogram**. The widget offers many option, so we will show here a slightly more complex set-up.
+
+We split the circle into months of the year. Next, we split the vertical axis by an attribute *Was cleaned*, meaning the outer circle shows cleaned dumpsites and the inner circle the ones with the trash. Finally, we set the color to *Tires [\%]* and stick with the default *Mean value*. This will display the average percentage of tires found at each month, split into sites that were cleaned and those that weren't. From the plot, we can see that the sites registered in January and November have a higher percentage of tires in the dumpsites. Possible, this has to do with the obligatory change to winter tires in November, resulting in a higher-than-average number of old tires dumped in the nature. There is no discernible difference between sites that were cleaned and those that were not.
+
+We have also selected the sites that were registered in November and were already cleaned. We passed the data to the Data Table, where we can inspect individual dumpsites.
+
+![](images/Spiralogram-Example.png)
 
 #### See also
 
-[Aggregate](aggregate.md)
+[Moving Transform](moving_transform_w.md)

doc/widgets/var.md

@@ -30,7 +30,7 @@ Using this widget, you can model the time series using VAR model.
 Example
 -------
 
-![](images/line-chart-ex1.png)
+![](images/LineChart-Example.png)
 
 #### See also