As far as I can tell, most researchers have their own ways of attempting an analysis, most learn through experience rather than formal training. One of the toughest aspects of any analysis can be the start-up. I think any contributions to this regard would be useful to budding researchers.
Mark, I don't know if this fully addresses what you mean by preliminary data analysis, versus possibly being a step earlier, but you might find this didactic chapter by me and Jill Lubansky helpful:
Malone, P. S., & Lubansky, J. B. (2012). Doing your homework: Preparing data for structural equation modeling. In R. Hoyle (ed.), Handbook of Structural Equation Modeling, pp.263-276. New York: Guilford.
We started at data entry and worked our way forward.
Mark, I don't know if this fully addresses what you mean by preliminary data analysis, versus possibly being a step earlier, but you might find this didactic chapter by me and Jill Lubansky helpful:
Malone, P. S., & Lubansky, J. B. (2012). Doing your homework: Preparing data for structural equation modeling. In R. Hoyle (ed.), Handbook of Structural Equation Modeling, pp.263-276. New York: Guilford.
We started at data entry and worked our way forward.
Hello Patrick, you nailed the question on the head, thats exactly what I meant to ask. I just want to be able to compare what different researchers do, the last time I performed an analysis, my supervisor did not let me progress without creating an analysis plan for the data with just a basic eyeballing of the physical data. I thought it was a tad hard, but only later did I realize how useful that approach was! Thank you so much for that reference!
Mark, although perhaps not exactly to your original point, I would add that if you have the opportunity to consult with an epidemiologist/biostatistician during the design stage, much anguish and frustration can be avoided, and more doors can be opened than you at first envisioned. But more to your question: in the same way that an embryonic physician is instructed to observe the patient carefully before laying on of hands and probing, I would suggest the first step is a careful examination of the data, determining how and when the data were collected, the target population and any biases present. Then a cleaning-up of the outliers, taking care to remove true outliers and retaining the oddities which do not conform to your a priori model! The data should largely dictate whether parametric or non-parametric analyses are appropriate, acknowledging that a data transform might be necessary to adjust for non-normality, or to allow comparison with previous studies. Just some initial thoughts before jumping in to serious SPSS territory.
.
Aleksandr, It's not online due to copyright. I can share it with you as an individual if you give me a means of direct contact.
Pat
- Badges
- Science topic