I have a population of 135 athletes, im seeing the frequency of a disease (stingers syndrome) in this population. This is the independant factor.
The possible risk factors are: Age, experience in years of sports, history of disease, symtoms; numbness, tingling, weakness etc.
age, experience are numerical
the rest are qualitative categorical
I want to see if these are risk factors for the disease and if they cause an increase/decrease in the frequency of occurence of diease (independant variable)
Is logistic regression analysis applicable for this? if not, then which test?
and how would, i apply this on SPSS?
In your case, constrasting exposed to unexposed cases can either be on the relative scale (i.e. prevalence ratio, risk ratio, rate ratio, hazard ratio) or on the absolute scale (i.e. prevalence difference, risk difference, rate difference).
You need to decide on which effect measure has a significant fit in answering your RQ's. This may be able to have a fit to the appropriate model to compute estimates of the effect measure and standard error (or confidence interval). You may want to employ the magnitude of the point estimate and random variability.
i would suggest you look at the attached paper which does what you want but in R not SPSS and this reference: Clinical Prediction Models: A Practical Approach to Development, Validation, and Updating | Ewout W. Steyerberg | download (b-ok.cc)
Best wishes, David Booth
CHAID analysis (chi-square-automatic-interaction-detection) is a method available in SPSS. Similar to BLR (binary logistic regression), CHAID models produce a "c-statistics" (model probability) that can be evaluated the ratio of sensitivity vs. specificity as the AUROC Curve (area under the receiver operating characteristic curve). In the standard heuristic, AUROC of 0.75 or above, is considered a good fitting model and above 0.9 is considered to be an a strong to excellent model.
The chief strength of CHAID, is that you can include a myriad of independent variable co-factors in an exploratory analysis in order to "rule out" insignificant co-variates (at a given experiment-wise threshold 90%?, 95%, 99%) in a single step in the model, and "cut to the chase" rather quickly. If you put in Age, Gender, Race-ethnicity, Elixhauser co-morbid conditions, Socio-economic as well as your clinical indicators flags (i.e. numbness, tingling, weakness, + etc. ) as independent predictors of Stingers Syndrome Yes/No, an exploratory predictive model of factors that significantly predict a higher (or lower) prevalence of Stingers Syndrome, can be developed. Probably the most powerful aspect of CHAID (in my opinion) is that the model methodology works very well on data sets that are much smaller in size than those required under the assumptions of the model for the BLR method.
My advice: use CHAID to explore potential independent predictors of Stingers Syndrome (Age, Gender, Race-ethnicity, socio-economic factors, insurance payer, clinical symptoms, seasonality: week-year, month-year) and then use BLR modeling with the independent factors that CHAID has identified as significant at the ideal alpha probability rejection threshold for your model (0.2 80%; 0.1 90%; 0.05 95%; 0.01 99%).
Scale level variables that reflect the physical status of the environment should also be controlled for such as but not limited to : atmospheric pressure (i.e. altitude, temperature, relative humidity, air quality indicators particulates, ozone levels, geo-location (latitude, longitude). I realize that this is likely beyond the scope of the present research, but these environmental factors are well known influential factors (like seasonal variation) that can have a statistically significant effect upon human physiology in sports medicine.
I hope this helps on your way of developing a solid predictive model of Stingers Syndrome, that controls for interactive effects of co-morbidities, demographics, socio-economic status, and ubiquitous seasonal variation.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS