Problem using Visual Studio 2008 Winforms ReportViewer

I have an application built using VS 2008 (3.5 fx). I build a series of
reports (using VS 2005 (sigh)) and am now tryiing to render them using a
Windows form and the ReportViewer control in the VS 2008 app. However I
always get the error message "The source of the report definition has not
been specified" when I execute the line
m_reportViewer.RefreshReport();
When I try to view the very same report in a VS 2005 app it renders without
error. I can also view the report using Report Manager and preview it in the
VS 2005 app I use to build it. Any suggestions?
Thx
Helen
--
Helen Warn, PhD
Agile Software Inc.
www.agile-soft.comNever mind. I added and removed the ReportViewer control a few times,
switching between Versino 8 & Version 9, and finally it worked :)
--
Helen Warn, PhD
Agile Software Inc.
www.agile-soft.com

Problem using SQL Server BI Development Studio

Hello

I'm using SQL Server BI Development Studio to create a mining structure to forecast time series analysis.
The process to create a model for time series has these steps:

Create a connection to the Database;

Create a source view. Create relationships between tables using the foreign keys;

Create the model with the mining structure. Define the fields that I won’t to predict and the algorithm properties aren’t changed.

Is something wrong in these steps, I missing something?

Other problem is the utilization of the model created earlier. In some tests that I made, the models don’t do forecasting upper then 20 steps. The problem could be the number of lines? But other data set with a large number of null values doesn’t have any problem predicting more than 20 steps using the time series algorithm.

Can someone help me?

Thanks in advance.

Jo?o Santos

<!--[if !vml]--><!--[endif]-->

I don't think there is a problem. The nature of the time series algorithm we use - Auto Regression Trees with Cross Predict (ARTXP) - is that predictions in the future can become unstable, since predictions become dependent on predictions. When we detect this instability, we stop producing output - this behavior can not be changed.

The dataset with large numbers of nulls is using some sort of missing value substitution which actually smooths the series - although you will have to determine the accuracy (usually more data = better).

We are considering allowing users to control the "instability" threshold in future versions.

|||The time series algorithm has a mechanism for detecting instability in predictions and stops returning predicted values when the variance exceeds a certain threshold. The point at which this happens is data-dependent - this explains the behavior you're seeing.