Longitudinal study long data format data repeated variables implementation

[Naomi Rose]

I am recently conducting a longitudinal data analysis by using mixed effects regression. I am using the long format data to do the analysis.

Since my data includes different questionnaires and some questionnaires (A) are answered every week while some are answered every month (B), when I change the data from a wide to a long format, should I leave the week that the participants did not answer (B) blank, or should I copy the last response I have across the months until they have a new response? What about the demographic variable (also measured every month)?


To make it clearer, here is an example:

Do I leave it blank if they do not have answer for that week?

Participant ID.	week of month	Questionnaire A response (measured every week)	Questionnaire B response (measured every month)	sex
1	1	1	1	male
1	2	2	-	-
1	3	3	-	-
1	4	1	-	-
2	1	2	3	female
2	2	3	-	-
2	3	1	-	-
2	4	2	-	-

Or should I copy the answer that they have from the first week of the month?

Participant ID.	week of month	Questionnaire A response (measured every week)	Questionnaire B response (measured every month)	sex
1	1	1	1	male
1	2	2	1	male
1	3	3	1	male
1	4	1	1	male
2	1	2	3	female
2	2	3	3	female
2	3	1	3	female
2	4	2	3	female

Or is there a better solution suggested?

[Carlo Lazzaro]

Naomi:
welcome to this forum.
I'd copy the answer that they have from the first week of the month; otherwise, due to the missing value(s), Stata will delete the whole observation(s).

[Naomi Rose]

Naomi:
welcome to this forum.
I'd copy the answer that they have from the first week of the month; otherwise, due to the missing value(s), Stata will delete the whole observation(s).

Thanks Carlo! If I copy the answer, then we are assuming the score doesn't change throughout that month. What if in reality the score changes, but it is not possible for us to measure it weekly since the questionnaire is done monthly?

[Carlo Lazzaro]

Naomi:
in fact, aftre the first and only measurement, you cannot say whether the monthly score changes or not throughout the month
Therefore, I'd assume that the monthly score does not change throughout that month in the baseline analysis and change this assumption in scenario(s) sensitivity analysis.

[Naomi Rose]

Naomi:
welcome to this forum.
I'd copy the answer that they have from the first week of the month; otherwise, due to the missing value(s), Stata will delete the whole observation(s).

Also, I thought mixed effect regression deal with missing values by MAR (missing at random)? Will they delete the whole observation?

[Naomi Rose]

Also, I thought mixed effect regression deal with missing values by MAR (missing at random)? Will they delete the whole observation?

Never mind I understand it. Thank you!

[Carlo Lazzaro]

Naomi:
this is not correct.
No Stata commands deals with missing values automatically.
In addition, Stata omits via listwise deletion each and every observation that has a missing value in any variable.