MVSUMM rolling medians

[Leonie Schneider]

I am trying to calculate rolling medians of stock returns every 3 days but i don't know how to use the mvsumm function. Any help would be greatly appreciated!

[Nick Cox]

mvsumm is a command, not a function, and it can be installed from SSC. (Posters are asked to explain the provenance of user-written programs they refer to: see FAQ Advice #12.)

I don't know what is proving difficult here as use of the median is a documented possibility.

I would typically use tsegen now, also from SSC, which differs slightly in by default using the data to the maximum extent possible.

Compare e.g.

webuse grunfeld

mvsumm invest, stat(median) win(3) gen(inv3ymed) 
 
tsegen inv3ymed2 = rowmedian(L(-1/1).invest)

[Leonie Schneider]

Thank for the response nick! Using the above code is giving me "variable date_5_min not found" error. I have no idea why STATA is referring to a variable called "date_5_min" as i have not mentioned it anywhere in the code

[Nick Cox]

Sorry, but there is nothing reproducible here in #3. You need to show any problem with the exact code you are using and a reproducible data example.

Do swing by

http://www.statalist.org/forums/help#stata (crucial)

and

http://www.statalist.org/forums/help#spelling (incidental).

[Leonie Schneider]

Sorry for not being clear earlier. I have a variable in my data called stock returns and in an attempt to generate a new variable "ret3med" which would show the rolling median of the past three returns, i used the following code:

mvsumm returns , stat(median) win(3) gen(ret3ymed)

This is giving me "variable date_5_min not found" error

Hope that was clearer!

[Nick Cox]

It is clearer, but I can see nothing informative about your data.

Did you read the FAQ Advice #12 as suggested?

Note that, as documented, mvsumm requires tsset (or xtset) data; if for some bizarre reason your time variable has been removed or removed since that setting, it is unlikely to proceed.