Chebyshev's Theorem -

any distribution
P(A1B) =
A 2nd B 1 75 % 20 = = 30 v 6
.
· xample =

2 88 9 % 30
75 % 30-2(b) 30 + 2(6)
.

to
.

3 93 75 % 40
PLAUB) Be p(a) p(B) P(A1B)
.

A + P(A1B) PAIB)
.
= =
or
-

P() =
1 =
P(a) P(B)
Empirical Rule-normal distribution (bell-shaped
# De
Morgan
P(a) =
p(anB) + P(an)
P(1B) =
1 -
P(AUB) .
1 68 % 10 of mean example mean80 . 5
3D
.

PrUB) =
1- P(AB) ou Pru) =
P() + P(B) -
P(n) 2 . 95 % 20 1073 -
85
P(BUA) . 99 3 % 30
3
=
1 -
p(b) .

20 70 -
90

Mutually exclusive PlanB) = 0 so PlAUB) = Pra) + P(B) and PranBend) = PrA) P(B) ... - 30 65 -
95

Independence PrA1B) =
P(A) P(B) and P(AIB) = P(A) and PrAuB)
-

pra) + p(B)(1-p(a)) and PrAIB) P(a) PLAIB) Types

=
= =
of sample
*
Bayes Theorem
I simple Random-no prior into about population
P(A/B) P(B/A) P(A) P(A) P(BIA) P(B(A) =
1 -

P(B/A)
=Y P(A/B)
= - =
-

P(B) I P(a) P(bla)] + [P(E)

·
·
Pipia)] P(B/) = 1 -

p(B/) estratified -
Prior knowledge of population to create Strete (subgroups)
/member) choosen , then
3 Systematie-rendom number
given interval

4 Cluster-mini populations , select someone 3

using I or

Sensitivity PCT + /Dt) Prevalence P(D )

+

> convenience-data readily available Biesed

- .
.

specificity P(T-ID) Positive Predictive P(D +/T + )

.
6 voluntary-self-selected

↑( False Positive) PCT + /DY) Negative Predictive P(DT) independent v

. (explanatory) -

manipulated

dependent /resulting) -

response
P(false Negative) P(T-ID + )
·
Interval -
No zero Jelook , ealender yr., RorF)
Discrete -
whole numbers Continuos -

integers Ratio-zero) Time elepsed , measurement , K)

Table
P(B1A) = P(A) P(B/A)
P(BIA)
.

①
A I
pra) P(B/A) P(B1A) = P(A) . P(B(A)
①
B P(BIA)-a V.- PrA))
. PIBIE) -

AV . .. Prs)
P(E)
Prbia) P(B1) = P() P(BI) .

⑧
Es
PrA)(A v)
-
. .

-
PrB/) P(BR) = Pr) P(BI) . A V
. .

(eatter plots , Ber graph ,

summarizing data
Descriptive and Circle
Organizing graph , histograms,
-

stems Leaf
, Frequency ( Seatter Plot
* positive
Interferential-Semples to draw inferences from population r = 1

L
P: correlation eeff. negative = -
pop .

Parameter Population -

stratified r:
semple . . .
Statistic-Semple
-

r coefficient determination
=

Census Measurements from entire r <0 25 no relationship

-
population .

. 25 1r10 5
0 week
Optimum
.

0 51r10 75 moderate
Random Sampling . .

r >0 75
.

strong
·
separate the
groups
Permutation -
order matters, repetition is not ellowed upr
mummuus 2
Population , sample, in Favor
,
Totel

~ infar
. Population same = X :
100 =
+
u ! >
for words total words sample
-

(M ! ) /we ! ) - repetition

Combinations-order doesn't metter , ner

mumm Histogram
- -
Split stem

Ist stem end stem

mmmus
class width =
largest-smellest leaves of 0-4
# elesses e-way : leaves 5-9

Ist 2nd 3rd 4th 5th

N = ·
Count how Midpoint =
Supper class limt lower class him)
many N
(2 , 3) (4 , 5) (6 , 71/8 , 9)
5-way : (0 , 1)
... in
-
2
u= &
Label and more than N
group reject any
Boundaries
or back-to beek : semple / I steml semple 2
.... / ...
) +elways add same amount ① consecutive intervals
to next (
02 ...
.. /. ... ) +
reject 2 2nd Low-ist upper

~ look now division , and subtract from tho 2

not rejected values 3 subtreet it from every low , and edd to

Last upper .
↑
Rejeet any repected number in the semple
chose as many asn asks
Fin
symmetric skewed left TFllT]] Bimodal Antti
777TT)
PS : when multiple 3
Labelling
IFTTTTTT unimodel symmG
no is possible ,
ignore
Uniform/Retangular
+
J- Distribution
 Fr
Eg
mean =
Ex
geometric meen = Ela, az .
W

median=Middle value ↓ variance =

SD
mode = most repected #

Popu Sam j2 = (xi x)2 -

S2 I
v 82

Range max-min ⑧ &

=
SD
32 E(xi - (2
E =

Midrange = min + max mean IH x -

1
2

Rule of
Thumb
=
-min
4 z value =
soore-mean
S D
..

min = X- 2 . SD
C V = SD
100 %
max =
.
. ·
+ 2 SD .

n
(25050075) = 0 , 9293
=
10 O

I
It upper IQR =
Q3 0. .
-

Lower
Step 1 =
(1 .
5)(IPR)
upper = 0 Step 1
.
3+

lower 1
0. -step
=

Experiments

·
Observational -

Don't manipulate one variable , casual relationship

Base Reports Detailed descriptions of the

-

diagnosis , contains
demographies
Case Control Case
, persons has disease in the
investigation
Control ,
group don't have disease , retrospective

Cross-sectional :
analyze data from a
single paint in time, compare affected and not

Conort studies Longitudinal, researchers raise questions about things

-

the disease. People with no disease are followed to see if they will
causing

·
Experiments change -
variables

single blind-patient doesn't know

↓
double-patient and researches

associated with
confounding variable -
both variables ,

distorts effect of variable