calculating variance and covariance using a minimum and maximum no. observations

[Batool]

Hello everyone,

I need to calculate stock beta using the formula: beta= covariance (Rm,Ri) / variance (Rm). I need to calculate beta using 24-60 observations. So, I need to create a new column that has the betas for each date-company calculated depending on the last 24-60 observations (no problem if it started from the current observation). I can create lags but I want more practical solution please giving that my data have several companies. here is how my data looks:

date                 company       Ri      Rm

Jan-2001              A                      3%       3.5%

Feb-2001              A                      3.2%     3.9%

Mar-2001              A                      2.7%     3.1%

.

.

.

Jan-2001              B                      3.9%       3.5%

Feb-2001              B                      4.1%     3.9%

Mar-2001              B                      3.3%     3.1%

.

.

your help is appreciated.

thank you in advance 

[PaigeMiller]

PROC CORR will calculate covariances and variances for you. Use the BY statement to have this calculated for each company.

[Batool]

thank you Paige Miller, I know that correlation can provide such information, however, the only thing is to determine the time span between 24 and 60. what I need is a macro with a loop but I do not know how to construct it.

[PaigeMiller]

---

@Batool wrote:

thank you Paige Miller, I know that correlation can provide such information, however, the only thing is to determine the time span between 24 and 60. what I need is a macro with a loop but I do not know how to construct it.

---

You do NOT need a macro.

Explain further, give a concrete example of what you mean by "determine the time span between 24 and 60".

[Batool]

i need 24-60 observations to calculate the  variance and covariance. so the minimum number of observations used to calculate the variance and covariance should be 24 and the maximum is 60. 

[PaigeMiller]

So if there are less than 24 observations, then you don't calculate the variance and covariance? Is that correct?

If there are more than 60 observations, then what do you do?

[Batool]

that is right. if there is more than 60 i want to use the last 60 only. 

I was searching and I think I found what i need (code attached),

can you please help me making it a roll regression so instead of calculating the beta using cov/var, i want to calculate it as the estimated coefficient from the model: (Ri-Rf)= beta (Ri-Rm) using the same window.

[PaigeMiller]

---

@Batool wrote:

that is right. if there is more than 60 i want to use the last 60 only. 

I was searching and I think I found what i need (code attached),

can you please help me making it a roll regression so instead of calculating the beta using cov/var, i want to calculate it as the estimated coefficient from the model: (Ri-Rf)= beta (Ri-Rm) using the same window.

---

So if I am interpreting this properly, you no longer need help with the original problem? Is that correct?

What is a "roll regression"? What is Rf?

[Batool]

I mean a macro regression. and Rf is another variable in my data.

[PaigeMiller]

So please define "macro regression".

[Batool]

I mean a window