Data Viewer | Vose Software

Data Viewer

See also: ModelRisk functions and windows, Presenting results introduction, Graphical descriptions of model outputs, Statistical descriptions of model outputs, The Bootstrap, VoseFrequency

 

 

 

ModelRisk incorporates a Data Viewer feature that allows you to quickly review your data sets prior to doing any analyses like distribution, copula or time series fitting. The Data Viewer is accessed by clicking on the Data Viewer icon:

which opens the following interface:

The data to be reviewed need to be held in a contiguous range. One proceeds as follows:

1. Select the range of cells in which the data are held in the Select Data range dialog

2. If you wish to see labels for each variable in the analyses (recommended) then tick the Labels box and select the location of the data labels

3. Select whether your data are for a single variable (Univariate) or several variables (Multivariate)

4. Select whether the data are simple observations (Data) or form a time series

5. The Copying options are only relevant if you wish to create an Excel report at the end of your review. Select ‘Live link’ if you wish to maintain a link to the original data (useful if you think that the values in your data set may be revised).

6. Click OK  

Depending on the options you have selected the Data Viewer will present your data differently. The four modes relate to the following combinations:

Univariate data set

Multivariate data set

Univariate time series

Multivariate time series

Each of these modes is described below.

Univariate data set

Clicking OK as above having selected ‘Data’ and ‘Univariate’ will display a window with two tabs. The data view tab shows a histogram of your data:

The Univariate Data Analysis tab gives a comprehensive graphical and statistical analysis of your data:

The left (red) and right (green) markers can be dragged across the graphs and the relevant data values and cumulative fractions are shown in the right Markers section of the statistics pane. One can also type in LowerX, UpperX, LowerP or UpperP values directly in this pane.

The scale of the horizontal axis can be changed by clicking this icon: .

The percentiles of the box plot can be changed by clicking .

Graphs can be copy/pasted to other applications by clicking .

The left hand pane provides reporting options if you wish to create an Excel report. Clicking the Create Report button creates a report with a large number of statistical analyses including non-parametric bootstrap assessments:

Multivariate data set

Clicking OK as above having selected ‘Data’ and ‘Multivariate’ will display a window with three tabs. The ‘Data view’ tab shows a histogram of each variable and scatter plots to visualize any correlation between the variables:

The ‘Points’ slider control allows you to vary the number of points in the scatter plots. This is useful if you have a lot of data because a scatter plot can get too blocked to show detail.

The Percentiles tick box will toggle between showing values or percentiles in the scatter plots. It is often easier to see correlation patterns for long-tailed variables if one uses percentiles.

The Logs tick box will toggle between showing values or log values in the histogram plots. It is often easier to visualize long-tailed variables if one uses logs.

Sliders can be used to split the data into groups. In the following example, a slider splits the Edge thickness variable at a value of 3.2:

The scatter plots then show those points in blue and red that correspond to this split.

Double-clicking on a histogram plot will show that variable in the Univariate Data analysis window, which is the window described above. Double-clicking a scatter plot will display the Multivariate Data analysis window:

This window provides information on each variable and the correlation structure between the two.

Selecting the Enlarge matrix option will swap the location between the bottom left correlation matrix and the scatter plot, helpful in better reviewing the correlation matrix if you have a large number of variables in the data set.

The correlation matrix displays correlation or covariance between each variable. The slider can be used to highlight correlations above a certain level. Selecting the |abs| option will allow you to highlight correlations whose absolute value is above some threshold.

Univariate time series data

Clicking OK as above having selected ‘Time Series’ and ‘Univariate’ will display a window with two tabs. The ‘Data view’ tab shows a time series of the variable. If you input a set of dates for the time series data they will be shown on the horizontal axis:

Moving the markers in the ‘Select times’ slider will allow you to ‘play’ the time series, meaning that the series will be presented as a video with the number of points shown controlled by the distance between these markers. If you are analyzing financial data it will generally be more useful to select the LogReturns option.

The second tab, ‘Univariate Time series’ allows you to analyze the data in a number of key ways, which can be switched on and off:

Autocorrelation shows a correlogram displaying how the variable is correlated with values over different numbers of lags. In the example above, there is no autocorrelation calculated to be significantly different from zero with (1-alpha) probability. Statistically significant autocorrelations are shown as red bars.

Moving average simply smoothes the data set by averaging over the defined number of periods.

Moving standard deviation calculates the standard deviation over the defined number of periods. It is useful to see whether there are periods of higher and lower random variation.

Linear regression fits a straight line to the series and reports the slope (b) and intercept (a) values on the graph and in the related fields.

Remove seasonality shows estimated seasonality factors and optionally removes these factors before the other analyses of the data.

LogReturns allows the user to analyse and graph the data in terms of the log return (essentially the proportional change from one period to the next). This is very useful for the prices of stocks and financial instruments but also for any other variable for which movement is likely to be proportional to its size (like population sized).

Multivariate time series data

Clicking OK as above having selected ‘Time Series’ and ‘Multivariate’ will display a window with three tabs. The ‘Data view’ tab shows a time series of each variable and scatter plots to visualize any correlation. If you input a set of dates for the time series data they will be shown on the horizontal axis:

Moving the markers in the ‘Select times’ slider will allow you to ‘play’ the time series, meaning that the series will be presented as a video with the number of points shown controlled by the distance between these markers. This is particularly useful to get a feel for whether a correlation structure is fixed or varies over time. If you are analyzing financial data it will generally be more useful to select the LogReturns option.

The second tab is the Univariate Time Series window, which is the same as described above.

The third tab is the Multivariate Time Series window, which plots any two variables together. Variable can be shown as a scatter plot:

or time line:

A new pair of variables can be selected by either double-clicking a scatter plot in the Data View tab, or clicking a cell in the correlation matrix.

 

Navigation