Biostats advice for neurosurgery projects

[Guest]

Anyone have good advice about where to start if you want to be able to do your own stats for projects? Would love any advice/ resources people have found useful. Tried a bit with R in the beginning of med school and found it complex.

[Guest]

SPSS is probably the most user-friendly program to start with, point and click interface with easy data management options. You can do all your basic stats for 90% of retrospective studies there (chi-square, T-test, ANOVA, regressions, KP curves, etc.) although it is definitely limited for more advanced things like trial statistics and various visualization options. 

SAS is also an option (similar but more Mac-friendly I've heard, haven't used it myself). 

If you seriously want to do high-end statistical research then the various packages and flexibility that R gives you is unparalleled (aside from Matlab or Python), so might be good to use SPSS while learning R on the side for the future.

[Guest]

Agree with above. SPSS is one of the most user friendly of the dedicated statistical programs. It's what I use for all my projects that require basic statistics. There are some extensions you can get to do more advanced statistics, but it requires some research and knowledge of python. A big limitation of SPSS I've found with a lot of the retrospective studies using large databases is the inability to propensity match. For these studies I have had to find someone who uses another program.

Going down the line of things that give you more flexibility at the cost of a higher learning curve, you also have JMP and Stata. As you've experienced, R is highly flexible but requires a lot of back end work to get what you want out of it.

In terms of the learning the statistics themselves, just start by reading other studies. Read the studies critically, think about the study design, groups they used, and variables they looked at, and then look at the statistics they did. Google the tests and decide if you think they were appropriate or not. Once you read about a handful of tests and when they should be appropriately used, you'll get a good grasp of the statistics lingo. Once you have your own database/study question, go through the same process of thinking, then use the statistical lingo you've learned to google and make sure the tests you're thinking of are appropriate.

[Guest]

Super helpful thank you both!