epidemiobiostats

Which statistical test will most likely be used and why?

* Matched pairs t-test: 1 nominal + 1 interval, linked

  • Husbands & wives: 1 nominal (2 non-inedependent groups)
  • Glucose ratio data: 1 interval

* Chi-square test: 2 nominal (2x3)

  • 1 nominal: Call sick VS. didn’t call sick
  • 1 nominal: 3 different shifts (3 groups)

* Pearson correlation: 2 interval

  • 1 interval: study time in hours
  • 1 interval: test scores

* t-test: 1 nominal + 1 interval

  • 1 nominal: Same-sex siblings VS non same-sex siblings (2 groups)
  • 1 interval: testosterone levels

* Two-way ANOVA: 1 interval + 2 nominal

  • 1 interval: time (faster VS not faster)
  • 1 nominal: trained VS not trained
  • 1 nominal; vacutianer VS standard syringe

* Chi-Square:  2 nominal

  • 1 nominal: margarine VS butter
  • 1 nominal: cancer VS not cancer

* One-way ANOVA: 1 nominal + 1 interval

  • 1 nominal: efficacy of qx in 3 groups (radical mastectomy VS radiation VS no tx)
  • 1 interval: quality of life measured on a 10 point scale.

*Chi-square: 2 nominals

  • 1 nominal: pass VS fail
  • 1 nominal: 3 test sites (3 groups)

* One-way ANOVA: 1 nominal + 1 interval 

  • 1 nominal: 3 test sites (3 groups)
  • 1 interval: test scores

* Matched pairs t-test: 1 nominal (linked, before and after) + 1 interval

  • 1 nominal: HTN pts 1w before & 3m after taking drug (2groups)
  • 1 interval: BP changes

Examples taken from Behavioral Sciences Kaplan Lecture Notes

Types of Mortality Rates

Regionville is a community of 100,000 persons. During 1985, there were 1,000 deaths from all causes. All cases of tuberculosis have been found, and they total 300: 200 males and 100 females. During 1985, there were 60 deaths from tuberculosis, 50 of them in males.

1.- The case fatality rate for tuberculosis is:
a. 6%
b. 20%
c. 2%
d. equal in males and females
e. cannot be computed from data given

2.- The crude mortality rate in Regionville is 
a. 300 per 100,000
b. 60 per 1,000
c. 10 per 1,000
d. 100 per 1,000
e. cannot be computed from data given

3.- The cause-specific mortality rate for tuberculosis is 
a. 60 per 100,000
b. 300 per 100,000
c. 200 per 1,000
d. 20%
e. cannot be computed from data given

Reliability: 

  • Consintency, precision.
  • Ability of a test to measure something consistently.
  • Threat: random error (is unfortunate but expected).

Validity

  • Accuracy.
  • Degree to which a test measures what was intended.
  • You need reliability to have validity.
  • Threat: systematic error (bad and biases result)
How would you calculate relative risk reduction (RRR)? [New drug vs standard care]

RRR = [Risk (control) - Risk (treatment)] / Risk (control)

Where:

  • Risk (control) = recurrence rate on standard therapy.
  • RRR = rate of disease recurrence after treatment.
  • Risk (treatment) = maximal incidence of the disease on standard therapy + new drug treatment.
youtube

Sensitivity, Specificity, Prevelance (by 100lyric)

Observational Studies and its Analysis MNEMONIC

Cross-Sectional Study: Chi-Square (x2)

  • cross in a square

Case-Control Study: Odds-Ratio

  • In case you didn’t notice, you can't control the odds.

Cohort Study: Absolute Risk & Relative Risk 

  • The ones left. I don’t have a mnemonic for this one…