Dropping variables that change in a longitudinal survey

[Roberto Villa]

I will try to be brief.

Race fixed effects have changed in my 4 different years. In the first survey some individuals that answered black for example changed to Asian Black in the year after.

How can I delete all these observations that show different answers in different years?
I just want to remove all the individuals that answered two different things in the survey

• For example the same Person (PersonNum=1109) answered: Race=101 in T1 , Race 102 in T2 , Race 101 in T3 .

or for example

• For example the same Person (PersonNum=20013) answered: Race=101 in T1 , Race=missing in T2 , Race 101 in T3

For both scenarios I need to drop them


Is there someone that can give me a quick guidance code, Im new with Stata

[Mike Lacy]

Assuming that the person and survey wave variables here are named "id" and "wave," one common way to do this is:

bysort id (race):  gen byte racevaries = (race[1] != race[_N])
drop if racevaries == 1  // will work, but not a great idea
// Better:
RunSomeAnalysis if (racevaries == 0)

Note that if an individual's race is missing in all waves, this variable won't catch it.

In Stata terminology, what is being done here involves identifying *observations* to be dropped, not *variables.* That clarification might help you in making sense of Stata documentation.

[Roberto Villa]

Sorry Mike Lacy but this also accounts for the following:

• For example the same Person (PersonNum=1109) answered: Race=101 in T1 , Race 102 in T2 , Race 101 in T3 .

or for example

• For example the same Person (PersonNum=20013) answered: Race=101 in T1 , Race=missing in T2 , Race 101 in T3

For both scenarios I need to drop them

[Nick Cox]

See the concurrent thread https://www.statalist.org/forums/for...-type-industry and its links.

[Mike Lacy]

Roberto Villa
Perhaps I misunderstood what you intended. In your second post, I'm guessing that you might mean that those two persons you described are *not* handled the way you want by the code I suggested. I'm confused because that code does mark them as having responses for race that vary across time.

. input int PersonNum byte T race
     Person~m         T       race
  1. 1109 1 101
  2. 1109 2 102
  3. 1109 3 101
  4. 20013 1 101
  5. 20013 2 .
  6. 20013 3 101
  7. end
. //
. bysort PersonNum (race):   gen byte racevaries = (race[1] != race[_N])

. list
     +--------------------------------+
     | Person~m   T   race   raceva~s |
     |--------------------------------|
  1. |     1109   1    101          1 |
  2. |     1109   3    101          1 |
  3. |     1109   2    102          1 |
  4. |    20013   1    101          1 |
  5. |    20013   3    101          1 |
     |--------------------------------|
  6. |    20013   2      .          1 |
     +--------------------------------+

[Roberto Villa]

Sorry Mike Lacy I appreciate your response , yes the


| Person~m T race raceva~s | |--------------------------------| 1. | 1109 1 101 1 | 2. | 1109 3 101 1 | 3. | 1109 2 102 1 | 4. | 20013 1 101 1 | 5. | 20013 3 101 1 | |--------------------------------| 6. | 20013 2 . 1


7 105 1 101 1
8. 105 2 101 1
9. 105 3 101 1

In this example , I just want to include observation 105 , I want to drop 1109 2013 , due to missing value and because of error in answer.


Thank you for your response


Roberto

[William Lisowski]

This question was reposted at

https://www.statalist.org/forums/for...-fixed-effects

with subsequent discussion.