Statistics in Med school and Beyond

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

ZorkDork1

Full Member
7+ Year Member
Joined
Dec 1, 2015
Messages
84
Reaction score
23
Hello, I had a question regarding the level of statistics knowledge required for success in MS and beyond.

I'm not exactly a wizard at math and statistics. Even when I did research as an undergrad, I did a lot of scut work, and didn't fully understand the statistical analysis that went into generating the graphs that I then talked about in poster presentations.

Think I should brush up on this beforehand? Or is it covered in medical school?
Or is it as simple as understanding p-values, figuring out if something is significant or not?

Thanks!
 
I'm trying to brush up on this before I start school. It is a really b**** to truly understand. I high suggest you read 'ending medical reversal.' This book is phenomenal in mapping out how to approach reading medical journals. However, this is more about reading articles rather than setting up studies.

Make sure you understand confidences intervals (CI) because p values don't tell you nearly as much.
 
Last edited:
Stats are a rabbit hole you could spend a career going down and still not explore it all. It can be as simple or as complex as you need it to be. Not sure you'd find much benefit brushing up ahead of time, though if someone knows of a good resource - the pathoma of stats - then that might be worth a peek at some point. Have no fear that you will be taught enough about stats, though you will always find things that are beyond you. And you will always find statisticians who like to quibble about them and make even seemingly simple things sound daunting.

Doing research is probably the best way to get a handle on it. Your reading and discussion with your statisticians will be far more meaningful than any lecture. If you find yourself really drawn to it, consider some advanced coursework in the future or with your MD like a masters or PhD.
 
Stats are a rabbit hole you could spend a career going down and still not explore it all. It can be as simple or as complex as you need it to be. Not sure you'd find much benefit brushing up ahead of time, though if someone knows of a good resource - the pathoma of stats - then that might be worth a peek at some point. Have no fear that you will be taught enough about stats, though you will always find things that are beyond you. And you will always find statisticians who like to quibble about them and make even seemingly simple things sound daunting.

Doing research is probably the best way to get a handle on it. Your reading and discussion with your statisticians will be far more meaningful than any lecture. If you find yourself really drawn to it, consider some advanced coursework in the future or with your MD like a masters or PhD.


Haha, the rabbit hole analogy is apt. I have little desire to tumble down it.

I guess my goal is to be in a place where I feel comfortable being part of a research team in MS and interpreting new developments in medicine as I begin practicing.

I guess anyone could look at a research paper, read the abstract, and say oh okay, this therapy is looking promising in treating disease X. But I want to be able to really look at the data and have a good idea how they messed with the numbers. In other words, I want to be a more critical reader and interpreter of science. In your experience, will med school emphasize this?
 
Haha, the rabbit hole analogy is apt. I have little desire to tumble down it.

I guess my goal is to be in a place where I feel comfortable being part of a research team in MS and interpreting new developments in medicine as I begin practicing.

I guess anyone could look at a research paper, read the abstract, and say oh okay, this therapy is looking promising in treating disease X. But I want to be able to really look at the data and have a good idea how they messed with the numbers. In other words, I want to be a more critical reader and interpreter of science. In your experience, will med school emphasize this?

Yeah medical schools in general don't emphasizes analyzing journal articles nearly enough. Hopefully there is a school that meets your needs. If your goal is getting good at looking at articles, that book I recommended above will give a ton of excellent insights.
 
Goljan's RR Pathology actually has a really good stats section for those interested.
 
Hello, I had a question regarding the level of statistics knowledge required for success in MS and beyond.

I'm not exactly a wizard at math and statistics. Even when I did research as an undergrad, I did a lot of scut work, and didn't fully understand the statistical analysis that went into generating the graphs that I then talked about in poster presentations.

Think I should brush up on this beforehand? Or is it covered in medical school?
Or is it as simple as understanding p-values, figuring out if something is significant or not?

Thanks!

I did statistics before medical school. Our school strongly emphasised evidence-based medicine. I thought the curriculum was comparable to the highlights of a lower division health statistics class. That's more than p-values and includes basic ideas like the odds ratio, relative risk, study designs, etc. (when I say basic, I don't mean easy--these are still subtle and tricky concepts).

Most research teams that deal with complicated data just get a statistician. No medical school is going to teach you, for example, generalized linear models or survival analysis. You can try to pick up statistical techniques as you go along, but that'll help more with interpreting other studies or having a rough idea of how to plan your own.

Statistics is like medicine. It requires a good foundation. A good foundation for statistics involves calculus, linear algebra, probability, and statistical programming at a minimum. And it's very easy to not know what you don't know. There are many, many published papers rife with amateurish statistical errors. And not only in insignificant journals.

You'd be surprised at how much "scut-work" a lead author has to do. And you'd also be surprised at how little of the analysis the lead (or tail) author actually understands.

Best way to learn is to do. Join a research team, reach out to the statistician, be aggressive about learning new techniques, and join a good journal club.

Also, this is going to be controversial, but I think writing skill, creativity, and tenacity matter so much more. You can always get a biostatistician. You can't get a bioenglishperson. That person has to be you.

Edit: One more thing, if you're looking to publish in medical school, there are lots of statistically unsophisticated (which is not the same as bad or unvalued) projects to work on, like meta-analyses or case series or case reports, etc. Meta-analysis is a tremendous skill to learn, because it teaches you to really scrutinise papers and scour the literature. Maybe you can do a Cochrane review? It's just seems unlikely that you'll be dealing with complicated data without the help of a statistician.
 
Last edited:
Statistics is like medicine. It requires a good foundation. A good foundation for statistics involves calculus, linear algebra, probability, and statistical programming at a minimum.
I couldn't agree with this more. It's unfortunate how many people think that "being good at statistics" is clicking some buttons in SPSS (or worse, MS Excel) and saying something is statistically significant (this goes way beyond medicine, too).

And it's very easy to not know what you don't know. There are many, many published papers rife with amateurish statistical errors. And not only in insignificant journals.
One of my favorite professors from grad school (a statistician) was telling me about many of the researchers he knows (non-statisticians), "They're very smart people. Incredibly smart. The problem arises when they know just enough statistics to get themselves into hot water without realizing it." As you said, there are a ton of iffy articles published in great journals. It leaves you scratching your head as to how they were accepted. What's worse is the number of people who will assume that it's a perfect paper because it's in a reputable journal (granted, it still could be a perfectly good paper, but there are many cases where it isn't).

For the OP: many people will tell you that you don't need it or it's not important, but if you ever plan to look into a treatment that a patient asks about, you need to have at least a basic understanding. If you want to do research, you're more valuable if you have some idea of what needs to be done (or can talk to a statistician without getting frazzled-- you speak medicine and a little stats, they speak stats and maybe a little medicine). If you can do an analysis properly, you'll be incredibly more valuable. If you're going to brush up on general stuff, get an undergraduate statistics textbook with worked examples. You can read and understand a book all day, but as lymphocyte said, you gotta do it to get your bearings and figure out how to handle unusual data sets or issues that arise in an analysis.
 
I would have a basic understanding of common statistical methods and what conclusions you can actually draw from them (and yeah, definitely understand what a p value means as well as what a standard deviation is, etc, and basic stuff like that).

If you're doing research, it's a good idea to have a fairly solid background to draw from so that you can interpret your results and present them correctly. If you can use R, that's awesome for research. If not, knowing the theory behind things like t-test, chi squared test, ANOVA, etc will help you expand upon them and use the pertinent methods for your own research.
 
😕😕😕😕😕 did you have any stats exposure prior to med school????

Yes, but you only need basic stats knowledge for med school exams. Just providing advice for OP, who said they are "not exactly a wizard at math and statistics."
 
Yes, but you only need basic stats knowledge for med school exams. Just providing advice for OP, who said they are "not exactly a wizard at math and statistics."
I realize I was short in that post - @lymphocyte 's post perfectly sums up why. I've had more exposure to stats than most med students, but the "you don't know what you don't know" phrase is very apt. Stats in med school (my experience), step review books, qbanks, and step itself, is treated pretty much like vocab - just memorize these terms and regurgitate (don't worry about actually understanding them)....hence my previous comment.
 
Last edited:
I realize I was short in that post - @lymphocyte 's post perfectly sums up why. I've had more exposure to stats than most med students, but the "you don't know what you don't know" phrase is very apt. Stats in med school (my experience), and step review books is treated pretty much like vocab - just memorize these terms and regurgitate (don't worry about actually understanding them)....hence my previous comment.
I'm in a similar position as you are with the stats background. I would agree that the med school stats (including what I've seen of review books) is basically non existent. As you said, it's vocab with no real understanding.

For the OP: I would try to understand both p-values and confidence intervals very well. They're both important and they're intertwined. If you realize the connection between them as well as what you can and can't say from each, you will be better off than most of your peers.

Aside from the basics of hypothesis testing and interpreting confidence intervals, I would recommend learning applied regression analysis (simple/multiple linear first and then simple/multiple logistic--much easier to understand the latter once you have a solid grasp on the former) and some survival analysis methods. Try to take note of when you're using a parametric vs. non-parametric technique (and why you're doing so).
 
Last edited by a moderator:
I'm in a similar position as you are with the stats background. I would agree that the med school stats (including what I've seen of review books) is basically non existent. As you said, it's vocab with no real understanding.

For the OP: I would try to understand both p-values and confidence intervals very well. They're both important and they're intertwined. If you realize the connection between them as well as what you can and can't say from each, you will be better off than most of your peers.

Aside from the basics of hypothesis testing and interpreting confidence intervals, I would recommend learning applied regression analysis (linear first and then logistic--much easier to understand the latter once you have a solid grasp on the former) and some survival analysis methods. Try to take note of when you're using a parametric vs. non-parametric technique (and why you're doing so).

All of the things you mentioned really fall into place with a solid understanding research methods and design, and that's what is really lacking in the stats required for and taught in med school. Physicians are expected to be solid critics of research, but with the way stats and research methods are taught in med school the only way you can learn the skills necessary to adequately do that (unless you have prior experience) is going out of your way to gain actual exposure to them.
 
Last edited:
All of the things you mentioned really fall into place with a solid understanding research methods and design, and that's what is really lacking in the stats required for and taught in med school. Physicians are expected to be solid critics of research, but with the way stats and research methods are taught in med school the only way you can learn the skills necessary to adequately do that (unless you have prior experience) is going out of your way to gain actual exposure to them.
The thing is, I think there is adequate time to teach some sort of foundation (at least from my experience). My school had a course dedicated to it, yet not much was taught in terms of experimental design and basic statistical concepts. With a revamp, the course could most certainly cover basic statistics, experimental design, critical appraisal of research, and even touch on the applied side of statistical analysis in the context of medicine and public health (i.e. how and when to conduct an ANOVA, verify the assumptions, and how to use a nonparametric technique if needed [violated assumptions, for example]). I guess it's also a function of those that are in charge of designing the courses at each school-- if they view it as unnecessary, you're not going to see it much.
 
If you can draw a 2x2 table and understand how to calculate sensitivity, specificity, NPV/PPV, an odds ratio, RRR, ARR, a likelihood ratio, and NNT/NNH, you know more than enough statistics to get through medical school.

If you can do the above as well as actually understand when/where the specific calculations are appropriate (and when they aren't), you know more statistics than most residents... in any field.

Edit: replaced ARV with ARR. Not sure where that typo came from.
 
Last edited:
If you can draw a 2x2 table and understand how to calculate sensitivity, specificity, NPV/PPV, an odds ratio, RRR, ARV, a likelihood ratio, and NNT/NNH, you know more than enough statistics to get through medical school.

If you can do the above as well as actually understand when/where the specific calculations are appropriate (and when they aren't), you know more statistics than most residents... in any field.
Both very true and very sad.
 
The problem with statistics is that it's not intuitive. Have you ever tried to explain probability to anyone? Just because something happened doesn't mean it had to happen, nor does it imply that was it the most likely outcome. If you can understand that, you understand more than 90% of people out there.
 
If you can draw a 2x2 table and understand how to calculate sensitivity, specificity, NPV/PPV, an odds ratio, RRR, ARV, a likelihood ratio, and NNT/NNH, you know more than enough statistics to get through medical school.

If you can do the above as well as actually understand when/where the specific calculations are appropriate (and when they aren't), you know more statistics than most residents... in any field.


I currently don't know how to do any of those... I'll get on that! Thanks!


Sent from my iPhone using SDN mobile
 
The problem with statistics is that it's not intuitive. Have you ever tried to explain probability to anyone? Just because something happened doesn't mean it had to happen, nor does it imply that was it the most likely outcome. If you can understand that, you understand more than 90% of people out there.
Statistics is often intuitive if you've studied it (based in logic, and uses mathematics and probability). Probability, is less intuitive than statistics (which, it is worth pointing out, that stats and probability are two different, but related, things).

Your point is received, though. I think that with qualified people teaching (mostly [bio]statisticians) or with students being required to take some form of statistics in undergraduate, that med schools could eliminate some of the discomfort and lack of intuition that many med students have regarding the topic.
 
bumping this thread. anyone have anything new to add in regards to this topic?

Good places to learn stats, Scope of knowledge required etc
 
bumping this thread. anyone have anything new to add in regards to this topic?

Good places to learn stats, Scope of knowledge required etc

Coursera is a good source. They list many different types of Stats courses

Basic Statistics Courses | Coursera

I participated in their Medical Neuroscience course for graduate/ medical professionals to review for Step 1. It was through Duke University and free.

Very helpful resource
 
Coursera is a good source. They list many different types of Stats courses

Basic Statistics Courses | Coursera

I participated in their Medical Neuroscience course for graduate/ medical professionals to review for Step 1. It was through Duke University and free.

Very helpful resource
Good call on Coursera. Just be sure to look into the credentials of the person who taught it. I've seen some of the courses taught by people with a PhD or MS in (bio)statistics-- these are the courses you want to take if possible. Believe it or not, it's really common for people without an actual stats degree to tell you the wrong thing for elementary concepts. Some of the most common mistakes taught by non-statisticians: a p-value is the probability of a Type I error, or is the probability that the results are real (really incorrect); that a specific confidence interval, say 2-10, has a 95% chance of being right (also pretty bad). Both of those things are incredibly wrong, but I've heard them frequently from people without a real stats background. You don't want to be learning much statistics from someone who's confused about basic concepts like that, especially when you're trying to learn for your own use with research or making treatment decisions in a few years.

Here's an example of a "safe" course: Statistical Reasoning for Public Health 2: Regression Methods - Johns Hopkins University | Coursera

and check out the professor's education on his CV:
John McGready, PhD

This is a good example of who you should feel confident learning from...
 
Last edited by a moderator:
Top