I want to investigate the impact of bacterial infection, which could occur at any time of disease course, on patient prognosis. How can I choose appropriate statistical methods?
I think you must first consider study design. As a suggestion, consider a fixed disease or surgical procedure. To take a simple example:
hip replacement surgery in otherwise well people aged 60-65 say.
using hospital records retrospectively, identify a group of patients who developed infection within 10 days post op, and another group who did not , compare rehab times, indices of movement, incidence of respiratory infection by standard normal dist statistics.
to do this study more generally in one could look at survival retropectively and compare survival stats between infected and non infected groups.
to study the effect of infections occurring at different times in disease course,you could just select restricted groups and analyse separately or do anaysis of variance with no infection, infection at x weeks, repeat infection etc .
a porspective study would be difficult in general but in particular disease categories might be possible.
antibiotic use would be a covariate in all this.
Your research proposal is simple, but analysis may be quite triky. If the infection can occur at any time, age and time on study may be related to this variable and taken into account (time-to event or survival analysis would be necessary, perhaps with a dummy variable to a proper bias control).
You can select particular disease and test the association of bacterial infections in that disease by using chi square test means make to groups one those having disease but not having secondary infection and another group with disease and bacterial infection and copare the outcome
To select statistical methods without having research hypothesis first is difficult. According to your idea, you can study
1. correlation: is having bacterial infection correlated to disease prognosis? how strong, what direction?
2. regression: could having bacterial infection predict disease prognosis?
3. differences: Is disease prognosis different between those who had infection and those who did have?
4 survival and so on.
But it all depends on your research question and hypothesis, as well as your data type. The use of parametrics or non parametrics tests depends on your data. If the assumptions were met, all could produce valid and meaningful result.
your question can only be answered if objective of the study is clearly known , no one help you out otherwise .
- Badges
- Science topic