TRIM - frequently asked questions

This list is updated in March 2003 for TRIM release 3.2.

FAQ TRIM

  1. Adding a new year to the data results in changes of indices in earlier years. Is that correct?
  2. Are there any requirements to the data?
  3. How can you see if the model fits?
  4. How should one assess the best fitting model in case of overdispersion and/or serial correlation?
  5. How to interpret the overall trend?
  6. How to interpret the standard errors of indices?
  7. Is it possible to name the categories within a covariate?
  8. Is it possible to open outputfiles generated in earlier TRIM runs?
  9. Is it possible to put all species data in one file and then assess indices for all species in one run?
  10. Is it possible to test the significance of change between two years?
  11. Is it possible to treat monthly and yearly data simultaneously?
  12. Is lack-of-fit of the model relevant in case there are few missing counts?
  13. Is there any lower limit in the number of sites to be meaningful? Not really, but few sites will usually result in large standard errors of the indices and trends.
  14. TRIM generates additive and multiplicative parameters. Which are to be preferred?
  15. TRIM generates model indices as well as imputed indices. Which are to be preferred?
  16. What are the advantages of incorporating covariates in the model?
  17. What do error messages mean as "model error: zero observations for changepoint at time 3"?
  18. What do error messages mean as "model error: zero observations for changepoint at time 3, category 1"?
  19. What do the error messages "invalid floating point operation", "singular matrix", "diverging parameters" mean?
  20. What does the error message "too many parameters" mean?
  21. What does the error message "unexpected value site." mean?
  22. What if the data are not Poisson distributed?
  23. What if the model does not fit?
  24. What is a covariate?
  25. What is overdispersion and what is its impact on the indices?
  26. What is the impact of serial correlation on the indices?
  27. What is the meaning of the 20-year period in Appendix C in the TRIM manual concerning the overall slope?
  28. What is the meaning of the stepwise procedure to select change points?
  29. What is the meaning of the Wald test for significance of deviations from linear trend?
  30. What is the use of the "no time effects model"?
  31. Which model should one use: the time-effects model or the linear trend model?
  32. Which proportion of missing values in the data is allowed?
  33. Which year should be chosen as the base time year?
  34. Why are the default p-values in the stepwise selection 0.2 and 0.15 instead of 0.05?
  35. Why has no bootstrapping being applied to assess standard errors of indices?