How to solve this confidence interval question ? - USMLE Forums
 USMLE Forums         Your Reliable USMLE Online Community     Members     Posts
 Home USMLE Articles USMLE News USMLE Polls USMLE Books USMLE Apps
 USMLE Forums How to solve this confidence interval question ?
 Register FAQs Members List Search Today's Posts Mark Forums Read

 USMLE Step 1 Forum USMLE Step 1 Discussion Forum: Let's talk about anything related to USMLE Step 1 exam

#1
05-11-2015
 USMLE Forums Veteran Steps History: Not yet Posts: 254 Threads: 33 Thanked 40 Times in 27 Posts Reputation: 50
How to solve this confidence interval question ?

A prospective study was conducted to assess the role of daily alcohol consumption in the occurrence of breast carcinoma. The investigator reported a 5 year relative risk of 1.4 for people who consume alcohol compare to those who do not.

The 95% confidence interval was 1.02 to 1.85. Which of the following 'p' values is most consistent with the results described above ?

A.0.04
B.0.06
C.0.09
D.0.11
E.0.20

Tell me how to solve this so confused.

I am trying to understand Confidence Interval and P value no where it explains properly.

Is it worth wasting my time on this for Step 1 ?

#2
05-11-2015
 USMLE Forums Scout Steps History: 1+CK+CS Posts: 66 Threads: 0 Thanked 101 Times in 50 Posts Reputation: 111

Quote:
 Originally Posted by ezguy A prospective study was conducted to assess the role of daily alcohol consumption in the occurrence of breast carcinoma. The investigator reported a 5 year relative risk of 1.4 for people who consume alcohol compare to those who do not. The 95% confidence interval was 1.02 to 1.85. Which of the following 'p' values is most consistent with the results described above ? A.0.04 B.0.06 C.0.09 D.0.11 E.0.20 Tell me how to solve this so confused. I am trying to understand Confidence Interval and P value no where it explains properly. Is it worth wasting my time on this for Step 1 ?
This is DEFINITELY worth your time for Step 1. Behavioral science, social science, and biostats make up 15% of your questions on step 1, but people rarely give it the attention it needs. It could sway your final score by 10-20 points or more.

So no, you are not wasting your time

There are a lot of picky definitions in stats, but I like to keep things simple, so at the risk of offending my fellow stats nerds out there:

p-value: this is the chance you got your results by pure chance. You know how you can flip a coin and have it come up on the same side 5-6 times in a row? That's just a sampling error and if you flip the coin enough times, you'll end up with a 50-50 split like you would expect. So in stats, the p-value just says what percent of the time your answer could be due to pure chance. For whatever reason, the field of science and medicine has decided that 0.05 (aka 5%) is the cutoff value for something being "statistically significant."

95% CI: This is the range of values where you are 95% sure the TRUE value of whatever you're measuring actually lies. For the example above it was Relative Risk, but you can construct of CI around just about anything. Obviously, the small the CI, the better because it means our result is more likely to be accurate.

**notice how the CI is 95% and the p-value cutoff is 5% --- that's not a coincidence**

Now, lets take a hypothetical and I'm going to ask you to calculate these values in your head.

Room A has 100 people in it; I measured ALL of them; their average height is 160cm with a standard deviation of 5 cm

Room B has 100 people in it; I measure ALL of them; their average height is 155cm with a standard deviation of 3cm.

From this I conclude that the people in room A are, on average, 5 cm taller than the people in Room B.

Calculate the p-value and the 95% CI. Go!
...
...
...
Time's up!

The answer is that there IS NO P-VALUE or CI!

In my example, I measured EVERYBODY. There's no sampling error to be had because I measured the entire population in question. Now if I had measured only 10 people in each room, I may or may not have found the same thing depending on who exactly I measured. In this case, I would have a p-value and a CI. As I measured more and more people in each room, the p-value and CI would get smaller and smaller until I measured everyone and then they would disappear. I would know "the truth."

So back to the original question at hand:
We're given a RR of 1.4 (95%CI 1.02 - 1.85) and asked to estimate the p-value. Remember that RR here means that people who consume EtOH are 1.4x more likely to get breast cancer. If the RR was 0.70, then people who consume EtOH would only be 0.7x as likely to get breast cancer; in other words, they would be LESS likely to get it.

So for Relative Risk, 1.0 is a big important number because that's where your stat completely changes its meaning.

Since I have a 95% CI that never crosses 1.0, I know that my RR has less than a 5% chance of being the complete opposite of what I'm saying (ie. more vs less likely of developing breast cancer).

Remember that p-value of less that 5% is significant?

So since my 95% CI never crosses 1.0, I know my p-value MUST be less than 5% or 0.05. Now all I have to do is pick the answer choice that is less than 0.05, so AA is the correct answer.

The faster ways to answer this one:
1) know that CI for RR or OR or HR that doesn't cross 1.0 is statistically significant; only one p-value on the list would be consider significant (A)

2) Use test strategy that says "which of these is not like the other?" Since only one p-value is significant and all the others are not AND they didn't give you enough information to actually calculate a p-value, you could assume they want you to pick the different one.

Hope this helps -- stats can be tricky!
 The above post was thanked by: CleverFOX (05-12-2015), ezguy (05-12-2015), gokulramani (05-28-2015), SRI101 (05-12-2015)
#3
05-12-2015
 USMLE Forums Veteran Steps History: Not yet Posts: 254 Threads: 33 Thanked 40 Times in 27 Posts Reputation: 50

Thanks so much for detailed explanation.

The 95% confidence interval was 1.02 to 1.85. you are not using 1.85 any where or 1.02.

#4
05-12-2015
 USMLE Forums Veteran Steps History: Not yet Posts: 254 Threads: 33 Thanked 40 Times in 27 Posts Reputation: 50

And how is RR and OR tied with confidence Interval ?

Quote:
#5
05-12-2015
 USMLE Forums Scout Steps History: 1+CK+CS Posts: 66 Threads: 0 Thanked 101 Times in 50 Posts Reputation: 111

Quote:
 Originally Posted by ezguy Thanks so much for detailed explanation. What about this ? this is confusing me The 95% confidence interval was 1.02 to 1.85. you are not using 1.85 any where or 1.02.
You can construct a Confidence Interval (CI) for almost ANY statistic out there: mean, Relative risk, odds ration, linear regression, correlation, etc. -- all of them can have a CI.

For ANY stat, all the CI is saying is this: I'm 95% certain that the TRUE value of whatever I calculated lies in this range of numbers.

In the question above, the actual numbers of the CI are not important. Only the fact that the interval does not include the number 1.0. This is what tells us the result is significant, and by our universally accepted definition of significance, the p-value must be less than 0.05.

Maybe this will help:

If the confidence interval in the question was 0.98 to 1.85, then BB would be the correct answer. In this hypothetical case, our CI would include the number 1.0. What 1.0 means in a RR is that there is not difference. If I'm 95% sure the RR could also mean there is no difference at all, then my p-value would have to be greater than 0.05.

Basically they just want you to understand that CI's that cross 1.0 are not significant.
#6
05-12-2015
 USMLE Forums Scout Steps History: 1+CK+CS Posts: 66 Threads: 0 Thanked 101 Times in 50 Posts Reputation: 111

Quote:
 Originally Posted by ezguy And how is RR and OR tied with confidence Interval ?
You can calculate a confidence interval for ANY statistic. Relative risk, odds ratio -- doesn't matter. I just listed examples of other stats where you can also see CIs. There are many others too.
#7
05-12-2015
 USMLE Forums Veteran Steps History: Not yet Posts: 254 Threads: 33 Thanked 40 Times in 27 Posts Reputation: 50

So since my 95% CI never crosses 1.0, I know my p-value MUST be less than 5% or 0.05. Now all I have to do is pick the answer choice that is less than 0.05, so AA is the correct answer.

it is more than 1.02 so it crossed 1 right ? Or I am i missing something ?

also what does not significant mean ? not important ?

Quote:
 Originally Posted by Young Doc You can calculate a confidence interval for ANY statistic. Relative risk, odds ratio -- doesn't matter. I just listed examples of other stats where you can also see CIs. There are many others too.

Last edited by ezguy; 05-12-2015 at 04:34 PM.
#8
05-12-2015
 USMLE Forums Scout Steps History: 1+CK+CS Posts: 66 Threads: 0 Thanked 101 Times in 50 Posts Reputation: 111

Quote:
 Originally Posted by ezguy So since my 95% CI never crosses 1.0, I know my p-value MUST be less than 5% or 0.05. Now all I have to do is pick the answer choice that is less than 0.05, so AA is the correct answer. it is more than 1.02 so it crossed 1 right ? Or I am i missing something ? also what does not significant mean ? not important ?
Nope, 1.02-1.85 is an interval that does not include 1.0. It gets close, but 1.02 > 1.00.

Significant is just an arbitrary term. It means that your results are most likely real and not just from random chance.

Message:
Options

## Register Now

In order to be able to post messages on the USMLE Forums forums, you must first register.
User Name:
Medical School
Choose "---" if you don't want to tell. AMG for US & Canadian medical schools. IMG for all other medical schools.
 AMG IMG ---
USMLE Steps History
What steps finished! Example: 1+CK+CS+3 = Passed Step 1, Step 2 CK, Step 2 CS, and Step 3.

Choose "---" if you don't want to tell.

 Not yet Step 1 Only CK Only CS Only 1 + CK 1 + CS 1+CK+CS CK+CS 1+CK+CS+3 ---
Favorite USMLE Books
 What USMLE books you really think are useful. Leave blank if you don't want to tell.
Location
 Where you live. Leave blank if you don't want to tell.

## Log-in

Human Verification

In order to verify that you are a human and not a spam bot, please enter the answer into the following box below based on the instructions contained in the graphic.