You are on page 1of 4

June 11, 2016

Kaplan Biostats:
Study designs
Observation
Passive collection of the data
No intervention
No control group
Data collected based on the timing
1. Past Data (retrospective)
case control study
start with Dx find factors
(Disease, exposures factors?)
Assess: odds ratio (OR)
CAUSATION
2. Present Data
Cross sectional study (collect the data
NOW)
factor Dx
Dx Factor
: I dont know which one comes first,
but just start analyzing.
Assess: square
Only CORRELATION, no
CAUSATION!

Experimental
Active collection of the data
Intervention
Control groups
1. Clinical trials
sample bias
2.Randomized Clinical Trial
measurement bias,
expectancy bias
3. Double blind randomized Control
Trial (RCT)
morals/ ethics
4. cross over DBRCT (Double Blind
Randomized Controlled Trial)
minimize the expentacy bias (-)

3. Future Data (prospective)


cohort study (observing, not analyzing)
factors Dx
: follow patients 30 years (if
someone develops lung cancer
obtain the factors from them)
Asses: Relative Risk (RR)
CAUSATION
In cohort study the following data is collected.
Disease
No Disease
Smoking
90
10
Non5
95
smoking
What is the odds that lung disease is due to smoking?
*But the study is cohort study not able to do the odds ratio!
Ans: cannot be determined!
OR : Case control study
Disease
No Disease
Exposure
A
B
No Exposure C
D
ODDS RATIO: (A/C)/ (B/D) AD/BC

RR: Cohort study


Disease
No Disease
Exposure
A
B
No Exposure C
D
RR: future incidence (new disease)
RR: I (E)/ I (NE) A/(A+B) / C/(C+D)
AR: I(E)- I (NE)
In cohort study: risk (RR or AR)
RR: times
AR: more ( I (E) : 10 and I (NE):2 AR = 10-2=8 8 more people got the disease
attributed to the disease!
Stats:
Descriptive

Inferential

Increase the sample size, more

inferential

Elements of standard
-Mean: sum / total number
-Median: middle number in series
-Mode: the number comes the most
1. Symmetrical distribution: mean=
median
2. Negatively skewed distribution:
mean <median
3. Positively skewed distribution:
mean > median
Q1) Timmy Jones USMLE score: 250
Mean: 220
SD: 15
What is his placable performance?
Mean
1 SD: 34
2SD: 13. 5
3SD: 2.5

1. Confidence interval (comparisons)


1) calculation
CI= mean +/- Z (SEM; standard error of
mean)
*clinical state: Z; 95% (2.0)
*statistical state: 99% (2.5)
2) compare
You look compare the intervals
There is overlap of intervals or no
overlaps.
If there is overlap, Difference that
exist is not significant
If there is no overlap, The
difference is significant!
3) risk (there is always a risk!)
CI
RR
Significant
0.9-2.0
0.65
No
0.6-0.9
0.45
CI<1, decreased
1.11.25
CI>1, increased
11.1

how many SD from the mean?


250-220 =30/15 =2 SD above the mean!-> he is in top 97.5 %.
____________________________________________
Q2) 100 women were weighed and the
mean weight was 120 lbs.
SD: 20
What is the CI @ 95%?

Does the CI contain value 1?


only the first one No significant difference
in Risk.

CI: if you jump down from the 5th floor, you


need larger size of the mat!
More degree of CI for the stunt man (who
only needs the small size of the mat) !

2. P value (Role of chance)


looking for a role of chance!
-5%= P= 0.05

Does the CI not contain value 1?


2nd and 3rd Significant difference in Risk.
CI >1: risk is significantly increased.
CI <1: Risk significantly decreased.

CI = mean +/- Z ( SD/n)


CI= 120 +/- 2.0 ( 20/10)
CI= 120 +/- 4 = 116 ~124 confident
interval

P > 0.05
; more than 5 out of 100 did not achieve
the result.
P< 0.05
; less than 5 out of 100 did not achieve the
result.
-To preserve reality;
xyz does not decrease the BP => H0
study result:
xyz decreases BP = HA
-extablished a significant difference from placebo
effect

1. P value

>0.05: 5/100 : => H0 is true!

(xyz did not lower the BP in >5/100 people)


correct decision:
Fail to reject the H0(null)
(accept the H0 null!)
investigator decides to reject the null H0
-concluding xyz lowers the BP
TYPE 1 ERROR
2. P value
<0.05: 5/100 : => H0 is true!

(xyz did not lower the BP in <5/100 people


correct decision:
Reject H0(null)
(establish HA)
-xyz lowers the BP
investigator decides to fail to reject the
null H0
-concluding xyz does not lower the BP
TYPE 2 ERROR

Q) in study of examining the effect of Lipitor on serum cholesterol. The investigator


establishes a H0 saying Lipitor does not lower the serum cholesterol.
The P value of his study is 0.07. He then recommended Lipitor as a drug to lower the
serum cholesterol.
He fails to reject the null. he made an error on type 1.
A.
B.
C.
D.

his recommendation is correct because Lipitor inhibits HmG reductase.


He is commiting a type 1 error.
He is commiting a type 2 error.
He is commiting a random error.

You might also like