Are you doing this within the chart itself? If so, then adding a linear trendline will create a regression line (line of best fit) only for all of the data points.
However, you could work around this by using several data series instead of just one. What you need to do is make the first subset one data series, and then make the next subset points a second data series. You chart will (after changing the marker types to be the same) look like there is just one series plotted, but you can then apply the trendline to just the first subset of the data.
Select the first subset in your data and create a Scatter chart. Now, right click in the chart and choose Select Data. You will get a dialog box that shows that you currently have one data series. Click the Add button to add another data series. You can specify a name (or not), and then choose the X values and Y values for this series. You want to select the second subset for the X values and the corresponding data for the Y values.
Now, you will have a different marker type for the first subset of points and the last subset of points. That makes it easy to distinguish the series on the chart, so leave it that way for now. Right click on the first series and choose Add Trendline. Make it linear and choose the option that places the equation on the chart. This will draw a line of best fit for only the first subset of the data points. If you want to, you can repeat this to get a second trendline for the second set of points.
Now, if you want it to look like there is only one data series, right click on one of them and format it so that it has the same marker type and color as the other.
Tim
http://feeds.feedburner.com/ExcelBlog