"SDN Average"

[the.ra.pist]

So I was interested in doing some data crunching on "SDN Average". Here's some facts about everyone who posted an exam review taken in 2013:

Total exams in data: 198
Average score: 248.93
Range: 206 - 275
Median: 250
IQR: 242 - 258
Mode: 249
STD: 12.64
SE: 0.90
CI 95% (true 'SDN Average' has a 95% chance of lying within this range): 247.13 - 250.73
Average +- 2 STDs (assuming no prior knowledge, a random person posting their score on SDN has a 95% chance of scoring in this range): 223.65 - 274.21
Most common exam month: June (97)
Least common exam month: December (0)
Best exam score month: February (254.8)
Worst exam score month: August (241.5)

Score progression across the year:

---

Members don't see this ad.

 

[thehundredthone]

What is the most likely conclusion can be drawn from this data?
A. Students should advance their exam dates to February.
B. The August scores are poor because of the "Upcoming changes to the USMLE Step1".
C. 5 brave souls took on the reporting bias of the SDN average.
D. People on SDN sometimes have way too much time on their hands.

With the amount of obsessing that happens over Step 1, I guess it's nice that there are finally some concrete numbers to point people towards.

 

[Promethean]

C and D

 

[DCBFan]

The national average that same year was a 227.

I'm of the (controversial? obvious?) opinion that the SDN reporting bias is a good thing. I don't think most of us come to this forum to get advice on how to score the national average.

 

[Hawked]

The national average that same year was a 227.

I'm of the (controversial? obvious?) opinion that the SDN reporting bias is a good thing. I don't think most of us come to this forum to get advice on how to score the national average.

The national average in 2013 was 228. They send the previous year average in the score reports so people commonly think that is the national average for their year.

 

[DermViser]

SDN has a huge selection bias (this is well established). SDN is not all people who take the exam in med school. It is not surprising that the SDN "average" is much higher than the national average (> 1 standard deviation). You wasted your time on that graph for nothing.

 

[GadRads]

SDN users with higher scores are more likely to report their scores herein. So the SDN average is definitely lower than 248.

 

[the.ra.pist]

SDN has a huge selection bias (this is well established). SDN is not all people is med school. It is not surprising that the SDN "average" is much higher than the national average (> 1 standard deviation). You wasted your time on that graph for nothing.

The point isn't to prove anything. It's simply to look at some stats about it broken down. I think I succeeded at that.

SDN users with higher scores are more likely to report their scores herein. So the SDN average is definitely lower than 248.

"SDN average" is the average score people self-report on the Step 1 thread.

 

[thehundredthone]

I love how even a thread about the SDN average cleared stated to be based on the 2013 experiences thread gets overanalysed.

 

[SBR249]

So I'm actually a little curious what the score distribution for this not-at-all-representative dataset looks like. The stats seem to indicate a slight left skew though not by too much.

SDN users with higher scores are more likely to report their scores herein. So the SDN average is definitely lower than 248.

To continue with the over-analyzing theme, it's not as clear cut as only high-scorers report their scores on SDN. There's also a discernible portion of people with low scores that self-report to seek advice or reassurance. This mitigates the upward shift in SDN average compared with the true population mean.

 

[the.ra.pist]

So I'm actually a little curious what the score distribution for this not-at-all-representative dataset looks like. The stats seem to indicate a slight left skew though not by too much.


To continue with the over-analyzing theme, it's not as clear cut as only high-scorers report their scores on SDN. There's also a discernible portion of people with low scores that self-report to seek advice or reassurance. This mitigates the upward shift in SDN average compared with the true population mean.

This graph isn't perfect, but I think it accurately represents the distribution:

As expected, there is a significant negative skew.

 

[chopes31]

Thanks for crunching it, this was fun to look at.


Posted using SDN Mobile

 

[Donald Juan]

Well, thank God I didn't take the exam in August.

 

[Th2]

SDN has a huge selection bias (this is well established). SDN is not all people who take the exam in med school. It is not surprising that the SDN "average" is much higher than the national average (> 1 standard deviation). You wasted your time on that graph for nothing.

...wut...​

that was the whole point of his stats. everyone "knows" the average here is higher. now we know. ty op,

this was fun to look at.

 

[SBR249]

Still kinda crazy that the reported SDN average is a whole std dev above the national mean. That means on average, an SDN score reporter is at approximately the 84th percentile.

 

[thehundredthone]

Still kinda crazy that the reported SDN average is a whole std dev above the national mean. That means on average, an SDN score reporter is at approximately the 84th percentile.

Why is it crazy? It's very reasonable to assume that that's the average person that would feel comfortable reporting his score on this forum. If n was higher, then things might be a little crazier.

 

[SBR249]

Why is it crazy? It's very reasonable to assume that that's the average person that would feel comfortable reporting his score on this forum. If n was higher, then things might be a little crazier.

I'm using it in a way that expresses admiration and esteem.

 

[autotroph]

Posted in wrong thread ><

 

[Psai]

Still kinda crazy that the reported SDN average is a whole std dev above the national mean. That means on average, an SDN score reporter is at approximately the 84th percentile.

Not really. If I got a lower score, I'd be more hesitant to post my score. Props to the people who post about their relatively lower scores when there are folks here sitting on 260, 270

 

[tantacles]

This is an awesome thread.

 

[thehundredthone]

I'm using it in a way that expresses admiration and esteem.

I do get that, and it's a good thing in general. But it's just a correlation, and self reported averages are always going to be higher. Now if using SDN bumped up all its members' scores, that'd really be something. The new Step 1 method would be UW + FA + bash everyone's score report for being "too high" on SDN.

 

[SBR249]

I do get that, and it's a good thing in general. But it's just a correlation, and self reported averages are always going to be higher. Now if using SDN bumped up all its members' scores, that'd really be something. The new Step 1 method would be UW + FA + bash everyone's score report for being "too high" on SDN.

 

[thehundredthone]

Posts like this pop up from time to time, some much less disparaging, but still bemusing.

 

[guitarguy23]

..

 

[dyspareunia]

So I was interested in doing some data crunching on "SDN Average". Here's some facts about everyone who posted an exam review taken in 2013:

Total exams in data: 198
Average score: 248.93
Range: 206-275
Median: 250
IQR: 242-258
Mode: 249
STD: 12.64
CI 95%: 223.65 - 274.21
Most common exam month: June (97)
Least common exam month: December (0)
Best exam score month: February (254.8)
Worst exam score month: August (241.5)

Your CI is wrong.

If the mean is 248.93, the sample size is 198, and the SD is 12.64, then the 95% CI is ~247 to 251.

I don't know how that affects your charts though.

 

[Th2]

Your CI is wrong.

If the mean is 248.93, the sample size is 198, and the SD is 12.64, then the 95% CI is ~247 to 251.

I don't know how that affects your charts though.

I think what he meant is that 95% of the population lies between 223 and 274, ie mean +/- 2 sd's. Whereas yeah, CI of the mean is 248.93 +/- 1.76, which is what you got.

 

[the.ra.pist]

You are both, indeed, correct. I don't believe any other data is tainted by that mistake. I fixed the post, thank you.

I admit to drinking while originally writing that post. Well, this post too. I'm starting to see a trend here.

 

[Th2]

 

[the.ra.pist]