Peer Firm Averages excluding ith observation

[Judith bo]

Hi All,
I am pretty new to Stata , so a bit lost on this one.
I am trying to find Peer Firm averages of Mean , Median and St.Dev grouped by Industry (SIC code) but excluding the ith (firm specific observation)

Any suggestions on how to approach this?

Thank you in Advance

Judith

[Ken Chui]

Welcome!

Please take a look of the FAQ here: https://www.statalist.org/forums/help. And pay attention on items 10-12 on how to craft an effective and efficient question, and then revise the current one by providing some example data and also more specific desired outcome. (For example, what is the condition of the i-th, etc.)

[Nick Cox]

Some of these problems are easy, but others benefit from focused code. The mean of others is at its simplest

sum of others / (sample size MINUS 1)

= (total of values MINUS this value) / (sample size MINUS 1)

with a twist if missing values are present

(total of values MINUS cond(missing(value), 0, value)) / (sample size MINUS !missing(value))

More generally, consider commands such as rangestat (SSC). and asrol (SSC).

. webuse grunfeld, clear

. sort year invest

. rangestat (median) invest , int(year 0 0) excludeself

. l invest invest_median if year == 1935

     +--------------------+
     | invest   invest_~n |
     |--------------------|
  1. |   2.54   33.099998 |
  2. |  12.93   33.099998 |
  3. |  20.36   33.099998 |
  4. |  24.43   33.099998 |
  5. |  26.63   33.099998 |
     |--------------------|
  6. |   33.1   26.629999 |
  7. |  39.68   26.629999 |
  8. |  40.29   26.629999 |
  9. |  209.9   26.629999 |
 10. |  317.6   26.629999 |
     +--------------------+

[Judith bo]

Hi Ken Chui and Nick Cox
Thank you for your help so far, I hope that this description is more precise

Here is an example of the Data

LPERMNO is the company code , each LPERMNO represents one firm,
I want to group the observations by fyear and SIC_1 excluding the firms observation and get the averages of those groups' mean median and st.dev

Output should look like this:

Thank you,

Judith

[Ken Chui]

Thanks for updating the question. We unfortunately cannot copy and paste a screenshot onto Stata, it'd be wonderful if you can look into the package "dataex" and provide some actual data example. That way users here can provide you with tailored codes that you can immediately use.

With the current first screenshot, I am still not sure if it illustrates what you need. Firm one has 8 data points, they are all under the same SIC 308, but their years are unique. Given that you said:

I want to group the observations by fyear and SIC_1 excluding the firms observation and get the averages of those groups' mean median and st.dev

1. You'll get the same data because the years are unique and hence there will be 8 year * 1 SIC combinations, which is still those 8 lines.
2. There is no exclusion from the above description. When an individual-level data are aggregated into summary, we don't usually refer the individual data being "excluded," they were just expressed differently.

Perhaps the next step would be i) if possible, check out dataex and provide us some sample data and ii) based on that data sample, tell us what you need summarized.