I am required to do a statistical analysis comparing Country A and Country B in terms of their approaches to minimize or prevent occurrence of COVID-19 in their respective country.
So far these are the information I have gathered:
- Positive cases from COVID-19
- Death cases from COVID-19
- Mitigation measures taken (Descriptive)
- Condition of country: health, economy, social
For now I can only decide to do T-test for death cases. Do you think it is suitable?
Feedbacks and ideas are very much appreciated. Thank you in advance!
Any meaningful statistical analysis aimed at comparing the efficiency of two countries in handling COVID pandemic MUST be based on dynamic data (daily or weekly) over a long period rather than on one-time data collected for one particular day, such as 14th August 2020. No meaningful analysis can be performed using only one-time data.
Hi,
is it correct that you have these numbers just once (meaning for instance total deaths region A vs.B) and not on a daily (or weekly) basis? It seems difficult to apply any methodology here apart from descriptive analysis.
Cristian Ramos-Vera : thought of that too, do you think one way or two is applicable in this case?
Kathrin Möllenhoff : yes, it's a one-time data collection. Meaning that I am to obtain these data (# of cases, death tolls. etc) on an online statistics database/website (e.g. for 14th august 2020 only).
Any meaningful statistical analysis aimed at comparing the efficiency of two countries in handling COVID pandemic MUST be based on dynamic data (daily or weekly) over a long period rather than on one-time data collected for one particular day, such as 14th August 2020. No meaningful analysis can be performed using only one-time data.
Indeed, that's what I meant as well. I don't see any meaningful analysis either.
Irisha H. If the pandemic was officially declared started on 15 March and you collect data up to the current day 14 Aug (5 months total) then, yes, it could be considered long-term dynamic one. After all, no more are available.
In this case, I believe, you could use the proportional hazards model.
It allows comparing survival over time between two different groups (conditions and/ or countries). You can include dummy variables to assess the importance of some factors or conditions (among others), say, factor G=1 for country A, G=0 for country B and so on.
The shortcoming of this approach is that you must test that the hazard functions ratio remains constant over time. Also, some variables should be indicated as censored at some time regarding the death case that is not happened yet.
On top of that, the cases and deaths should be stratified by the population groups: age, morbidity pre-existing conditions, healthy / unhealthy lifestyle, etc.
As for mitigation measures taken, the so-called Peltzman effect should be taken into account. This suggests that introducing one safety measure, such as, e.g. car seatbelts may lead to other, compensatory risk behaviours, such as speeding. (If you perceive that your car is safer than usual, you might compensate for this by driving faster.) In the context of COVID-19 the wearing of a mask may make people feel safer, and hence minimize other protective behaviours, such as social distancing and regular hand washing.
The bottom line: without taking into account multiple factors included into a validated dynamic model it is difficult to compare convincingly the efficiency between two countries in handling COVID-19 pandemic. No single statistical test (be it the t-test or ANOVA or any other) will be sufficient to do so.
Kathrin Möllenhoff : yes, sorry about that. I misunderstood your question, it isn't a one-time data collection.
Alexander Kolker : really appreciate the feedbacks and recommendations! Will look up these model and theory you mentioned. Hence no statistical test that is suitable for this case, is it? Since my study requires the use of at least one statistical analysis, but I am yet to discover a fitting one.
Irisha H. I believe that you have to focus on the methodology called survival analysis. The Cox proportional hazard model that I have mentioned above is an example of such an approach. You could consult a good book: Glantz, S. Slinker, B., Primer of Applied regression & Analysis of Variance, chapter 12, Survival Analysis and the Proportional Hazards Model, 2-nd Edition, McGraw Hill, 2001
Irisha H. complementing other answers, social sciences has a long history of policy evaluation, and based on that, I would suggest the following: (1) comparing each policies of each country (easy), (2) choose a specific policy you want to test (kind of easy), and (3) create a counterfactual to compare the output (e.g. CV19 cases) if the policy was not implemented, compared to the situation observed, controlling for confounding factors (hardest part, thousands of papers and books about it). Then, (4) you can compare the estimates of the policy impact obtained in (3) for each country, to see which one is worst/better, or if there is any difference at all (kind of easy too, but depends on the quality of the work made on step 3). In (3) different types of quasi-experimental models can be used to create the counterfactual situation (as e.g. PSM, AIPWs, diffs-in-diffs, discontinuity designs, etc), but these models should be properly extended with dynamic epidemiological models. Sounds like a beautiful -but very defiant- scientific project.
@Irisha H.
You can apply t-test,chi-square test,ANOVA test, Z-score test for proof the difference between the countries. But before run any test you have to know which test will be applicable on your data. for that purpose you have to know conditions or assumptions of individual test. If your data meets these conditions, you can run...
Irisha H. , I just saw a pre-print of @Marta Blangiardo that maybe of interest to you:
Preprint Estimating weekly excess mortality at subnational level in I...
Basically one additional approach to your problem would be to compare the excess mortality between the two countries, using the methods describe in the paper above.
Maybe of interest to you, also a recent paper of mine with a similar Bayesian spatio-temporal epidemiological methodology:
Preprint Environmental risk factors of airborne viral transmission: H...
- Badges
- Science topic