`plotpred` (Series)

Visualize multi-step-ahead prediction

Syntax

[Hx, Hy, Hm] = plotpred(~Range, X, Y, ...)

Input arguments

X [ tseries ]

Input data with time series observations.

Y [ tseries ]

Prediction data arranged as described below; the prediction data returned from a Kalman filter can be used, see Example below.

Range=Inf [ numeric | Inf ]

Date range on which the input data will be plotted.

Output arguments

Hx [ numeric ]

Handles to a line object showing the time series observations (the first column, X, in the input data).

Hy [ numeric ]

Handles to line objects showing the Kalman filter predictions (the second and further columns, Y, in the input data).

Hm [ numeric ]

Handles to one-point line objects displaying a marker at the start of each line.

Options

Connect=true [ true | false ]

Connect the prediction lines, Y, with the corresponding observation in X.

FirstMarker='None' [ 'None' | char ]

Type of marker displayed at the start of each prediction line.

HandleVisibility={'on', 'on', 'on'} [ cellstr ]

Visibility of handles to the lines created; the first element sets the visibility for the first line Hx, the second element sets the visibility for for the prediction lines Hy and the third element sets the visibility of the starting point markers, Hm.

ShowNaNLines=true [ true | false ]

Show or remove lines with whose starting points are NaN (missing observations).

See help on plot and on the built-in function plot for options available.

Description

The input data Y need to be a multicolumn time series (tseries object), with one-step-ahead predictions x(t|t-1) in the first column, two-step-ahead predictions x(t|t-2) in the second column, and so on. Note the timing assumptions.

If x1 is a series with one-step-ahead predictions x(t+1|t), x2 is a series with two-step-ahead predictions x(t+2|t), and so on, while x is a series with the actual observations x(t), the following command will create a time series that can be then passed into plotpred( ):

p = [ x1{-1}, x2{-2}, ..., xn{-n} ];
plotpred(x, p);

Examples

The plotpred( ) function can be used with prediction-step data returned from a Kalman filter, filter. The prediction-step data need to be specifically requested using the 'output=' option (as they are not included in the output database by default), with the prediction horizon assigned in the 'ahead=' option (the horizon is 1 by default):

[~, g] = filter(m, d, startDate:endDate, ...
    'output=', 'pred', 'meanOnly=', true, 'ahead=', 8); 
figure( );
plotpred(startdate:enddate, d.x, g.pred.x);

plotpred (Series)