OK - doing this helped for the Time Series project, so I'm going to start off again.

Here's my idea of what we have to do if you've chosen the Dummy Variables project:

1. Run through the examples starting on Page 20 of the project template found in: http://www.neas-seminars.com/discussions/shwmessage.aspx?ForumID=171&MessageID=5014. This should make you familiar with how inflation affects the residual plot. Essentially you should test with large discrete changes to the inflation rate (this should cause a V-shape in the residual plot), constant (continuous) inflation (this should result in a smooth curve), and increase in the stochasticity ie - sigma (random fluctuations should "obscure" the residual plot - in english, this means that the y-values will be more dispersed at each x-value).

[NEAS: Discrete change ÿ V-shape or upside-down V-shape

Continuous change

ÿ parabolic shape

Use a low

σ, such as 1%, to see the expected shape of the residual plot. Use a higher σ, such as 10%, 20%, or 50%, to see the effects of stochasticity.]

2. Throw in dummy variables - at this point you should make the decision of whether or not you are going to use a discrete change or continuous change inflation rate (choose only one), we should only use low sigmas. Your analysis will differ if you choose discrete or continuous, each are explained in the above project template on pages 24 and 25 respectively.

[NEAS: If the residual plot has a V-shape or an upside-down V-shape, use a dummy variable of D × (X₂ – k), where k is the value of X₂ at the apex or nadir of the V.]

3. After your dummy variable analysis you should have another residual plot (of course you will need to do the regressions with the Excel add-in) and if all worked correctly the plot should have three lines (residuals vs x₁, residuals vs x₂, and residuals vs x₃) all horizontal. You should probably do some verification that your plots are correct using the methods described on page 21 of the project template.

[NEAS: Plot against X₁ and X₂.]

4. Write-up - follow the criteria outlined in the project template on page 26.

[NEAS: The write-up is important; the SOA wants to see that candidates understand how to use the statistical concepts.]

Joey,

Can you clarify what you suggest to do in step 2.  Specifically what do you mean by "Throw in Dummy Variables"?

 

Thanks

On page 24 & 25 of the project template posted by NEAS, it describes what to do depending on whether you choose a continuous or discrete change in inflation.

For example, if you choose a discrete change, you will need to introduce a dummy variable into your regression formula (this is shown on page 24 "The regression equation is...".

Use that as your starting point and then you make adjustments at low stochasticity to your beta's.  Once you get a "feel" for what is supposed to happen, then you adjust them to more realistic figures (I'm still trying to figure this out).

JR

[NEAS: Start with σ = 0.01 to see what is expected. Then change σ to 10%, 20%, or 50%, depending on the type of project, the values of the geometric decay and the inflation rate, and the number of observations.]

Is everyone running this on the stable rates or unstable?

If you take a look closer they are both the same, with some modifications of the least square variables.

I don't know for sure, but I don't think it matters.  Choose one and do the analysis.  I think what I have found out from the TS project is that NEAS doesn't care so much about what you do, but as long as you demonstrate that you have learned the core concepts of the course and have demonstrated some proficiency in statiscal software (i.e. Excel).

Hope that helps.

JR

[NEAS: Some candidates want specific instructions, such as "Do steps A, B, C, …" The SOA does not want to see if you can follow instructions. The SOA wants to see if you can apply the statistical techniques to data sets without specific instructions.]