problem multiple observations panel data

[Josephine Bijl]

Dear everyone,

I am currently new at working with Stata, and currently working on my thesis with which I use panel data.

In my data I would like to measure how life events of people and their peers are related to an outcome on a wellbeing scale. My problem is that I have several observations per year.
Because of this, Stata is not able to recognize my data as paneldata. I've read that I can transform my data from long to wide. However, since I have a lot of years, I would like to avoid this. I can have up to 15 different observations per year, and I also included up to 15 different years, so that would make my data quite inconvenient to work with.

I read that it is maybe also possible to create a new variable, I would like to have some advice on this.

To make it more clear:
For every 1 ID, I can have observations for over 15 years. Within one year, I have 3 different persons that can have a possible life event: self/partner/family member. There are 5 different type of life events that can occur.

I was thinking about taking out the data in another file, generating 15 dummy variables, where I combine every 3x5 possible situation (person+life event1, person + life event 2, ....., partner + life event 1, partner + life event 2... ) and merging it back into the data again, combining it for ID and year, so that I would have all the possible observations on 1 row.

Thank you for your advice!!

To get an idea, a scope of the data:


input long ID int year long(event_who2 event_what2) float CESDtot
61100002 1996 2 4 65
61100002 1996 2 2 65
61100002 2002 2 4 61
61100002 2002 2 2 61
61100002 2005 3 4 70
61100002 2008 . 0 73
61100002 2012 3 18 74
61100003 1996 2 2 66
61100003 1999 . 0 64
61100003 2002 4 4 62
61100003 2005 . 0 65
61100003 2008 2 4 63
61100003 2008 2 2 63
61100003 2012 4 4 60
61100005 1996 2 2 47
61100007 1996 2 4 67
61100007 1999 . 0 76
61100007 2002 4 18 75
61100007 2005 . 0 76
61100009 1999 . 0 74
61100009 2002 2 2 68
61100009 2005 . 0 73
61100009 2008 . 0 73
61100009 2012 2 2 68
61100009 2012 4 4 68
61100010 1996 . 0 72
61100010 1999 . 0 65
61100010 2005 3 4 65
61100010 2005 2 2 65
61100010 2005 3 18 65
61100010 2005 2 17 65
61100010 2005 4 18 65
61100011 2005 . 0 61
61100011 2008 . 0 60
61100011 2012 . 0 63
61100012 1999 . 0 68
61100012 2005 4 4 75
61100013 1996 . 0 46
61100013 1999 . 0 46
61100013 2005 . 0 48
61100014 1999 . 0 68
61100014 2005 3 19 62
61100014 2008 . 0 70


end
label values event_who2 event_who2
label def event_who2 2 "Member of family", modify
label def event_who2 3 "Spouse or partner", modify
label def event_who2 4 "Youself", modify
label values event_what2 event_what2
label def event_what2 2 "Death", modify
label def event_what2 4 "Hospitalisation", modify
label def event_what2 17 "Separation or divorce", modify
label def event_what2 18 "Termination of activity or retirement", modify
label def event_what2 19 "Unemployment", modify
[/CODE]

[Carlo Lazzaro]

Josephine:
welcome to the list.
If the order of the observations is irrelevant, you can simply -xtset- your data without a -timevar-, that is:

xtset ID

See -help xtset- and related entry in Stata .pdf manual for more details.

[Josephine Bijl]

Dear Carlo,

Thank you for your reply.
Unfortunately, I also want to take into account how my well-being variable evolves over time after a shock, so I would need to take the order into account.

Best,

Josephine

[Carlo Lazzaro]

Josephine:
another recipe would imply -collapse-.
If that were the option, you should relabel your events (i.e. hospitalization + death for the same ID) and calculate a mean value for CESDtot per ID per year.

[Josephine Bijl]

Thank you for your help, I was also thinking about using collapse (I just created the dummy variables for each unique combination), but shouldn't that mean that I will lose a lot of other information such as gender etc?

[Carlo Lazzaro]

Josephine:
not necessarily so, as you can include -gender- between brackets after -by-.

[Josephine Bijl]

thank you Carlo, I tried to do it now, but for some reason it says that my variable is missing.

I created the dummy's and then ran the command:

collapse (sum) < new vars> by ID year CESDtot gender

Sorry for all my questions, I hope you can help me with this?

[Rich Goldstein]

please, as requested in the FAQ, post exactly what you typed and exactly what Stata responded; assuming that what you show in #7 is approximately correct (it can't be fully correct), you need to follow the instructions in the help file and place your "by" variables inside parentheses

[Carlo Lazzaro]

Josephine:
I do share Rich's point.
What you typed, as it is, cannot be completely correct.
Being positively in replying with such a scant handful of inputs requires a huge (and, in all likelihood, unuseful) guess-work.

[Josephine Bijl]

I am sorry that I did not put the code on the post clearly. I will definitely do so in the future. I wanted to let you know that I was now able to use the collapse command.
Thank you so much for your help!