Disease Detectives B/C

Test your knowledge of various Science Olympiad events.
Private Wang Fire
Member
Member
Posts: 100
Joined: June 1st, 2015, 3:43 pm
Division: Grad
State: OH
Has thanked: 0
Been thanked: 0

Re: Disease Detectives B/C

Post by Private Wang Fire »

Unome wrote:
1. If a variable is associated with both the exposure being tested and the disease/treatment being studied, but is not causally related to the exposure, then it is a confounding variable, or confounder.
2. Stratify or match the participants by age group, or just restrict the study to a certain age group (though the second may create further confounders).
3. Not sure exactly what's being asked; is this asking for a type of statistical test?
Cochran Mantel-Haenszel (statistical test for stratified data) (idk how to calculate it or use tho lol)
You got next?
MASON HIGH SCHOOL '18
User avatar
Unome
Moderator
Moderator
Posts: 4338
Joined: January 26th, 2014, 12:48 pm
Division: Grad
State: GA
Has thanked: 235 times
Been thanked: 85 times

Re: Disease Detectives B/C

Post by Unome »

Does that count as a test statistic? I always thought of those as being like Fishers exact or chi-squared, and Cochran-Mantel-Haenszel being a method of finding risk. And calculating it is iirc somewhat like a weighted mean using the total population of each stratum (I know this because my partner BS'ed it at the last tournament and got full credit - unfortunately her attempt at BS'ing a chi-squared problem didn't work as well :geek: )
An investigator is collecting data on an outbreak of botulism in the town of Milford in southern Delaware. The investigator suspects that a steak joint in the town may have been transmitting the disease. The investigator determines that of the 12 living cases (3 cases have died, and the investigator couldn't find information about them) 7 of them can remember having eaten at the steak joint. Using as a control a sample of 18 people randomly selected from Sussex County (the southernmost of Delaware's three counties), the investigator determines that 2 of those people remember having eaten at the steak joint.

1. What type of study is this?
2. Give three potential sources of bias in the study, and explain how each of them affects the data.
3. Choose one of the sources of bias that you listed above and explain how the study should have been conducted to mitigate that bias.

Assuming that the study was conducted accurately (without systematic biases) and the same data was collected:

4. Choose an appropriate measure of risk, explain why the measure is appropriate, and calculate it (show work).
5. Choose an appropriate measure of statistical significance, explain why the measure is appropriate, and calculate it (show work).
Userpage

Opinions expressed on this site are not official; the only place for official rules changes and FAQs is soinc.org.
yang573
Member
Member
Posts: 35
Joined: April 1st, 2014, 12:25 pm
Division: C
State: NY
Has thanked: 0
Been thanked: 0

Re: Disease Detectives B/C

Post by yang573 »

Yeah, I'd call CMH RR and OR as a form of analysis/adjustment rather than a statistical test.
1. case-control
2. rumination bias - exposure of controls may be underrepresented; Neyman's bias - dead not included, possibly underrepresenting the exposed group; interviewer bias - nature of questions asked may lead to overrepresentation of the exposed group
3. Neyman's bias: find a medium

4. Odd ratio is used because incidence can't be calculated. The actual size of the exposed group is unknown
OR = ad/bc = [(7)*(18-2)] / [(12-7)*(2)] = 11.2
5. I think a z-test would do, because you're working with proportions rather than numbers. I believe you're only required to understand statistics and not expected to calculate them. :P
My question is, how does a steak joint transmit botulism? Canned beef?
So much to do, so little time.
maxxxxx
Member
Member
Posts: 284
Joined: November 30th, 2015, 8:11 am
Division: Grad
State: PA
Has thanked: 0
Been thanked: 1 time

Re: Disease Detectives B/C

Post by maxxxxx »

yang573 wrote: I think a z-test would do, because you're working with proportions rather than numbers. I believe you're only required to understand statistics and not expected to calculate them. :P
On the statistics handout from soinc.org it lists the equation for z tests, so I'd assume we're expected to know how to calculate it. I'm pretty sure the only ones we're just supposed to understand are CMH, Fisher's Exact, and McNemar's.
Lower Merion Class Of 2017
yang573
Member
Member
Posts: 35
Joined: April 1st, 2014, 12:25 pm
Division: C
State: NY
Has thanked: 0
Been thanked: 0

Re: Disease Detectives B/C

Post by yang573 »

maxxxxx wrote:
yang573 wrote: I think a z-test would do, because you're working with proportions rather than numbers. I believe you're only required to understand statistics and not expected to calculate them. :P
On the statistics handout from soinc.org it lists the equation for z tests, so I'd assume we're expected to know how to calculate it. I'm pretty sure the only ones we're just supposed to understand are CMH, Fisher's Exact, and McNemar's.
Oddly enough, I've never been expected to complete a statistical test, but I've been asked to perform adjustments. At states last year, I had to do a CMH-RR.
Assume the conditions have been met.
H0: p = 1/2
HA: p > 1/2
significance level: 0.05

SD(p-average) = sqrt( (p-exact * q-exact) = sqrt( [(1/2) * (1/2)] / 30 ) = 0.00617284

z = (p-average - p-exact) / SD(p-average) = ((7/9) - (1/2)) / 0.00617284 = 3.042903097
p of z = 0.001172
p < 0.05
Reject the null hypothesis. The proportion of cases among exposed is greater than 1/2.
I'm not sure I trust those calculations.

While working on this problem, I found this interesting read: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2938757/ I'm not sure if a z-test is appropriate with a contingency table; a confidence interval may be more suitable

Thinking about it, I think a Fisher's Exact Test would also work.
Assume conditions are met.
H0: a/c = b/d
HA: a/c =/= b/d
significance level: 0.05

p = [(a+b)! * (c+d)! * (a+c)! * (b+d)!] / [a! * b! * c! * d! * n!] = [9! * 21! * 12! * 18!] / [7! * 2! * 5! * 16! * 30!] = 0.008469611

p < 0.05
Reject the null hypothesis. There is an association between the exposure status and disease status.
So much to do, so little time.
User avatar
Unome
Moderator
Moderator
Posts: 4338
Joined: January 26th, 2014, 12:48 pm
Division: Grad
State: GA
Has thanked: 235 times
Been thanked: 85 times

Re: Disease Detectives B/C

Post by Unome »

yang573 wrote:
maxxxxx wrote:
yang573 wrote: I think a z-test would do, because you're working with proportions rather than numbers. I believe you're only required to understand statistics and not expected to calculate them. :P
On the statistics handout from soinc.org it lists the equation for z tests, so I'd assume we're expected to know how to calculate it. I'm pretty sure the only ones we're just supposed to understand are CMH, Fisher's Exact, and McNemar's.
Oddly enough, I've never been expected to complete a statistical test, but I've been asked to perform adjustments. At states last year, I had to do a CMH-RR.
Assume the conditions have been met.
H0: p = 1/2
HA: p > 1/2
significance level: 0.05

SD(p-average) = sqrt( (p-exact * q-exact) = sqrt( [(1/2) * (1/2)] / 30 ) = 0.00617284

z = (p-average - p-exact) / SD(p-average) = ((7/9) - (1/2)) / 0.00617284 = 3.042903097
p of z = 0.001172
p < 0.05
Reject the null hypothesis. The proportion of cases among exposed is greater than 1/2.
I'm not sure I trust those calculations.

While working on this problem, I found this interesting read: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2938757/ I'm not sure if a z-test is appropriate with a contingency table; a confidence interval may be more suitable

Thinking about it, I think a Fisher's Exact Test would also work.
Assume conditions are met.
H0: a/c = b/d
HA: a/c =/= b/d
significance level: 0.05

p = [(a+b)! * (c+d)! * (a+c)! * (b+d)!] / [a! * b! * c! * d! * n!] = [9! * 21! * 12! * 18!] / [7! * 2! * 5! * 16! * 30!] = 0.008469611

p < 0.05
Reject the null hypothesis. There is an association between the exposure status and disease status.
Yeah Fisher's Exact was what I was going for, although I'm really not sure whether that makes sense. And I have no idea how one could get botulism from cooked steak, I just made up the scenario randomly without putting too much effort into a realistic scenario. Anyway, your turn!
Userpage

Opinions expressed on this site are not official; the only place for official rules changes and FAQs is soinc.org.
yang573
Member
Member
Posts: 35
Joined: April 1st, 2014, 12:25 pm
Division: C
State: NY
Has thanked: 0
Been thanked: 0

Re: Disease Detectives B/C

Post by yang573 »

A slightly less ambitious question:

Foodborne illnesses are often characterized by _____. List three.
So much to do, so little time.
User avatar
Unome
Moderator
Moderator
Posts: 4338
Joined: January 26th, 2014, 12:48 pm
Division: Grad
State: GA
Has thanked: 235 times
Been thanked: 85 times

Re: Disease Detectives B/C

Post by Unome »

- a single source of an infectious agent (though this is more applicable to outbreaks rather than the infection per se)
- the presence of toxins (not sure if this is actually a thing, but it seems pretty common)
- improper food preparation (again, more for outbreaks, but I can't think of anything else)
Userpage

Opinions expressed on this site are not official; the only place for official rules changes and FAQs is soinc.org.
yang573
Member
Member
Posts: 35
Joined: April 1st, 2014, 12:25 pm
Division: C
State: NY
Has thanked: 0
Been thanked: 0

Re: Disease Detectives B/C

Post by yang573 »

Unome wrote:
- a single source of an infectious agent (though this is more applicable to outbreaks rather than the infection per se)
- the presence of toxins (not sure if this is actually a thing, but it seems pretty common)
- improper food preparation (again, more for outbreaks, but I can't think of anything else)
That works. I guess it wasn't that good of a question. I was thinking of short incubation periods, gastrointestinal symptoms, and point-source outbreaks (which you said)
So much to do, so little time.
UTF-8 U+6211 U+662F
Exalted Member
Exalted Member
Posts: 1597
Joined: January 18th, 2015, 7:42 am
Division: C
State: PA
Has thanked: 6 times
Been thanked: 15 times

Re: Disease Detectives B/C

Post by UTF-8 U+6211 U+662F »

Reviving: What are the three things that characterize an agent's potential to spread and cause disease?
Locked

Return to “2017 Question Marathons”

Who is online

Users browsing this forum: No registered users and 3 guests