Kind of a difference in differences model

[Kjell Weyde]

Hi!

I am investigating whether people who experience a reduction in cycling time to a certain place will also show an increase in cycling, between Time 1 and Time 2.

I have the following dataset:
DaysCycled_it is the number of days cycled previous week at Time 1 and Time 2 for each participant;
Cycling_minutes_saved_it is the reduction (in absolute values) from Time 1 to Time 2 in minutes in cycling time from residence to a specific location. This will have 0 for everybody at Time 1, and a value corresponding to the number of minutes saved at Time 2.
Time is coded 0 for Time 1 and 1 for Time 2.

Which of the following models makes the most sense to fit, and how should I do this in Stata?

1) Mixed model:

 
 DaysCycledit = Cycling_minutes_savedit + Time + (1 | subject_ID)

2) Linear regression:

DaysCycledit=2 - DaysCycledit=1 =Cycling_minutes_savedit + DaysCycledit=1

3) None of these, and if so, how to specify a "correct" one? Yours,
Kjell Vegard

[George Ford]

Is minutes saved relative to driving (or some other transportation modality)?

[Kjell Weyde]

Minutes saved is not relative to other transportation modalities. If the ride took you 50 minutes at time 1 and 46 minutes at time 2, the person has a value of 4 for this variable.

[George Ford]

The more you ride, the faster you ride. So why isn't minutes saved a function of days cycled?

[Kjell Weyde]

Sorry, I should have specified that the cycling time is calculated based on Google Maps (could have used cycling distance instead). The change has come about because some changes in the infrastructure. So the cycling time is not related to any physical aspects of the participants.

[George Ford]

what is "saved"? or is time just time?

[Kjell Weyde]

I mean that for some, depending on where they live, to cycle to the same location at time 2 will take less time than it did at time 1.

[George Ford]

why?