Biostats

[hcrunner]

---

Members do not see this ad.

Hey, I was wondering if anyone could give a suggestion of a good Biostats book that is brief and gives the basic knowledge of the subject that someone needs for the USMLE? I really haven’t had any exposure to stats so I kind of just need the basics to get me through the test. Thanks and good luck to everyone taking the test this year

 

[Kashue]

Hey, I was wondering if anyone could give a suggestion of a good Biostats book that is brief and gives the basic knowledge of the subject that someone needs for the USMLE? I really haven’t had any exposure to stats so I kind of just need the basics to get me through the test. Thanks and good luck to everyone taking the test this year

hy biostats

 

[Pinner Doc]

I used HY Behavioral Science. The final two chapers are Biostats. I would think this is sufficient, combined with FA.

Any dissenters? I'm not very strong in Biostats, but I was hoping this would be enough.

 

[cherrypie]

I used HY Behavioral Science. The final two chapers are Biostats. I would think this is sufficient, combined with FA.

Any dissenters? I'm not very strong in Biostats, but I was hoping this would be enough.

I would look at hy biostats. If you dont want to read the book (really short) then at least read the chapters on sesitiviyt, specificity, etc.. He does a real good job of explaining this and it is much more thorough than hy behav science. You may or may not get lots of stat questions. But, if you do get lots on your exam and you do well then this can really boost your score. I had friends who did well on most topics of step 1, but the breakdown on their score sheets showed that the majority of the questions they missed were in the "beha science" section. So, if you know this topic well it can boost your score. good luck

 

[cdpiano27]

Sensitivity - the probability that you predicted a correct decision when the decision was indeed correct.

Specificity - the probability that you predicted an incorrect decision when the decision was indeed incorrect.

Ideally we want BOTH a high sensitivity and high specificity.

In categorical data, this corresponds to the following situation:

Y (response)
Correct Incorrect

(explanatory varable)
X correct Pr(1|1)= specificity Pr(2|1)

incorrect Pr(1|2) Pr (2|2)= sensitivity


I am not an osteopathic student at all but I am a PhD student in biostat who just finished the masters and this is how I learned it in my categorical data class. Maybe the COMLEX has different notations but this should be the main idea.

 

[cdpiano27]

Sorry, reversed Pr(1|1) = sensitivty!

and Pr(2|2) = specificity!

 

[cdpiano27]

One more thing important.

These numbers are used to graph the ROC curve which measures the area under the curve. We plot the sensitivity on the Y axis and 1 - specificty on the X axis. Hence to have the best curve we need high sensitivty and low 1 - specificity which is high specificity.

The c-hat value on the SAS printout measures how good this curve is and this in either proc genmod or logistic (forget which one) with option / ctable.

C ranges from 0 to 1 and the closer to 1 the better!

The c-hat value is used instead of the gamma value (ranges between -1 and 1) which measures ordinal association betwen y and x in certain situations by using concordant versus discordant pairs: for example a low number of tables but a high number of cell counts or the reverse: a high number of tables but a low number of cell counts. This creates bias in measuring the association between Y and X and we have to resort to different measures. The sensitivity / specificity is one way of doing this.

 

[Pinner Doc]

OMG, I *hate* this crap. I signed on to be a doctor, not a statistical analyst. I'll hire my own if I need one. Till then, can't we leave this stuff for the Public Health students?

 

[cherrypie]

OMG, I *hate* this crap. I signed on to be a doctor, not a statistical analyst. I'll hire my own if I need one. Till then, can't we leave this stuff for the Public Health students?

dont worry about the stuff posted in other messages. Just read through the big topics in high yield stats. He gives a very good simple explanation and he really does help you understand sens/spec. Rarely on step 1 will you be asked about the definition of sens/spec you must understand how and what the values mean. Be able to reword or define sens/spec beyond the basic definitions that we are all taught. The author had a very intuitive approach to stats. So, its worth your time to read some of it atleast. However, there is quite a bit that you can skip in the book. If you dont want to read the book, at least read his bullet points. Those are money. good luck.

 

[cherrypie]

One more thing important.

These numbers are used to graph the ROC curve which measures the area under the curve. We plot the sensitivity on the Y axis and 1 - specificty on the X axis. Hence to have the best curve we need high sensitivty and low 1 - specificity which is high specificity.

The c-hat value on the SAS printout measures how good this curve is and this in either proc genmod or logistic (forget which one) with option / ctable.

C ranges from 0 to 1 and the closer to 1 the better!

The c-hat value is used instead of the gamma value (ranges between -1 and 1) which measures ordinal association betwen y and x in certain situations by using concordant versus discordant pairs: for example a low number of tables but a high number of cell counts or the reverse: a high number of tables but a low number of cell counts. This creates bias in measuring the association between Y and X and we have to resort to different measures. The sensitivity / specificity is one way of doing this.

this is very impressive. But, you definately do not need to go this far for step 1 purposes. Although, with your background it will definately help you. Good luck.