| Temporal methods |
|
|
|
|
| ARIMA |
SAS 9.2 |
3 weeks |
00:01 |
Significant amount of testing needed for the final model inputs; also, it is a method that has not frequently been used at DOHMH so there was a learning curve; its description in the literature [19] and actual coding of the method were straightforward |
| C2 |
SAS 9.2 |
1 week |
00:01 |
One of the most commonly used methods; easy to understand; easy to code and well documented in the literature [26] |
| CUSUM |
SAS 9.2 |
1 week |
00:01 |
Commonly used method; determining inputs to the CUSUM model was the most difficult part; otherwise, easy to code and well documented [14] |
| GLM |
SAS 9.2 |
3 days |
00:01 |
Commonly used model; accessible to most analysts to understand, code, and troubleshoot [23] |
| Holt-Winters |
SAS 9.2 |
2 weeks |
00:01 |
Experience was similar to the ARIMA where significant time went into developing the model specifications; the method has not been frequently used but is accessible and not difficult to understand [23] |
| Temporal scan statistic |
SaTScan |
3 days |
00:04 |
Much of the programming is done already in the SaTScan program [27], so analysts only needed to define parameters; if called from SAS or R, will need to write programs to define the parameters and run macros |
| Spatial/spatio-temporal methods |
|
|
|
|
| GLMM |
R |
1 week |
0:50 |
Moderately difficult to program the model and output, knowledge of regression models needed [21] |
| Bayesian |
R, WinBUGS |
10 weeks |
2:00 |
Highly difficult; extensive knowledge of several statistical packages and advanced statistics is required to understand and implement [22,30] |
| Spatial scan statistic |
SaTScan, R |
3 days |
0:06 |
Much of the programming is done already in the SaTScan software [27], so analyst only needed to define parameters; some knowledge of spatial epidemiology is needed |
| Space-time permutation |
SaTScan, R |
3 days |
0:12 |
Same as spatial scan statistic [27] |
