Hi all,
I am wondering if anyone can shed some light on how to deal with my time-varying covariates.
As a side note: I am using STATA MP version 15.1
Specifically, I originally had a dataset with the following form
However, I now believe this is not correct since my covariates (gender, age and nationality diversity are time-varying, it is measured on a quarterly basis)
So I have changed my dataset to the following form (example using dataex).
BvdIdNumber: company ID
Gender, Age, Nationality: diversity indices for the group (Blau & coefficient of variation)
_st: 1 if observation is to be used
_d:1 if firm failed
_t: Analysis time, I now have it always for every quarter (0-1, 1-2...).
_t0: Start analysis tile
However, I am unclear how to deal with the violation of the PH assumption, as well as how to adapt the dataset.
Here are my main questions/concerns?
- I have an observation per quarter, but perhaps this is not ideal.
- Are my covariates time-varying? I believe they are, however they don't constantly change over time (eg. sometimes they change, sometimes they don't)
- I used tvc, but it might be better to use stsplit? If so at what time should I split (whenever one of the diversity variables is different to the previous period?)
- How to move forward after PH assumption violation for specifically time-varying covariates?
Thank you in advance for your advice!
Best regards,
Laura
I am wondering if anyone can shed some light on how to deal with my time-varying covariates.
As a side note: I am using STATA MP version 15.1
Specifically, I originally had a dataset with the following form
| Firm ID | Firm Failure | Quarters Survived |
| 1 | 1 | 3 |
| 2 | 0 | 40 |
So I have changed my dataset to the following form (example using dataex).
BvdIdNumber: company ID
Gender, Age, Nationality: diversity indices for the group (Blau & coefficient of variation)
_st: 1 if observation is to be used
_d:1 if firm failed
_t: Analysis time, I now have it always for every quarter (0-1, 1-2...).
_t0: Start analysis tile
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input str16 BvdIdNumber float(Gender Age Nationality) byte(_st _d _t _t0) "AT9070350951" 0 .25851214 0 1 0 1 0 "AT9070350951" 0 .25851214 0 1 0 2 1 "AT9070350951" 0 .2530698 0 1 0 3 2 "AT9070350951" 0 .2530698 0 1 0 4 3 "AT9070350951" 0 .2530698 0 1 0 5 4 "AT9070350951" 0 .2530698 0 1 0 6 5 "AT9070350951" 0 .24785185 0 1 0 7 6 "AT9070350951" 0 .24785185 0 1 0 8 7 "AT9070350951" 0 .24785185 0 1 0 9 8 "AT9070350951" 0 .24785185 0 1 0 10 9 "AT9070350951" 0 .24284475 0 1 0 11 10 "AT9070350951" 0 .24284475 0 1 0 12 11 "AT9070350951" 0 .24284475 0 1 0 13 12 "AT9070350951" 0 .24284475 0 1 0 14 13 "AT9070350951" 0 .23803593 0 1 0 15 14 "AT9070350951" 0 .23803593 0 1 0 16 15 "AT9070350951" 0 .23803593 0 1 0 17 16 "AT9070350951" 0 .23803593 0 1 0 18 17 "AT9070350951" 0 .23341388 0 1 0 19 18 "AT9070350951" 0 .23341388 0 1 0 20 19 "AT9070350951" 0 .23341388 0 1 1 21 20 "AT9070422953" 0 0 0 1 0 1 0 "AT9070422953" 0 0 0 1 0 2 1 "AT9070422953" 0 .0288615 0 1 0 3 2 "AT9070422953" 0 .05656854 0 1 0 4 3 "AT9070422953" 0 .05656854 0 1 0 5 4 "AT9070422953" 0 .05656854 0 1 0 6 5 "AT9070422953" 0 0 0 1 0 7 6 "AT9070422953" 0 0 0 1 1 8 7 end
Here are my main questions/concerns?
- I have an observation per quarter, but perhaps this is not ideal.
- Are my covariates time-varying? I believe they are, however they don't constantly change over time (eg. sometimes they change, sometimes they don't)
- I used tvc, but it might be better to use stsplit? If so at what time should I split (whenever one of the diversity variables is different to the previous period?)
- How to move forward after PH assumption violation for specifically time-varying covariates?
Thank you in advance for your advice!
Best regards,
Laura
