> yx{`!mET
j lS`:;xmRkA~MM6M+(MEZz6CJ0[
ZsMqtq7ިBbPI}tsNSt'ߝ# /?jB9~`!r5OD5g,`\@xڥTKSQ~]ܦ)ג4P(8HAsU#_92!𡗞z?C{
zߊ=D{weHxsB5tC
!jYPP53d5d@2Vh%@tn(}?WS?FDa{fSlbs_ЎD\A8%32Dg!@җϨ+4E@>2\HhN0ty`NEGurqmo:upm31 L7Z+Q3i+#E#&t'5ը0ӲxB
[U)*J'&I>3)XRs/;kfzV,{II`!].
hQ*".s h0.x͚k\Uyo~mԎm?t:DC^OSn
!%M?HYt5\꠸\h+нzlmoo'IQ6͌~#yX~y'Ɉ̫VVVꖈ;~ "QBCWiD\J2
l/70T7JcQףv:C>bn/xZ^bſSlRS`Nq]3rc{:;+S5eH$yLY`sϰ,'Y[Ǵ_=&jtBPl2)Oz5)e䒿)OlODQ]{,`t聠m5E<\Wp\
0=0˻=}so;sof1zC@swzuG遡 K>[f=JQ쟃m8#NF,8p,Yq(sTD4sNlziuhM=۾s?m`IF#r=me6>BDzt)qwˈU8QKýÙMGpps8dg:7A͓O.7OiwR?<#;oD'w;x[fsFtdĮ<`%"Ƿ3N3_31m&}}ӹ+{!\yD֔ qkb[fprӯgU^9g^U rB{4EzEd+"=H딶`h:Ru4rV>'OdO+ܑSV>""XDp^`įHF#eWAvNZWY[Xӱi[]%TWpE+zՓ'9RTzjVOh4"G5k5['k UOr4q/;:xlBZ#+d;R>m UO!\CS8B慦qX=p}M`QwVqΝ,+eh"G8LEЁC2YcLsC&~˻
:U1HGs8cXGsCHwtHNӹCGwt`z;<ya:7t>B_ECht0CX'F`sf8'gk!zKIx8JꊝJvi#W3)vi\̝4J.=w[Gѿ
m`!G0IdR;o&
P2xZoU?Avl hb$o@2@b,t K.ُ
}"n&fj`/M'/ATw;wwiM9seT&DhcB>L>g(}T\5jW^=\YO{6巳A:m,uf=V*jm%4Ig
0F,zoہv
̑G=A,{1Hl+*F]ޅѬm bht4rc\ނ[}@wyJ[Nokb@k&3NcP̱E=+NkNךa=G>1Cz
MKp
'Z7d)eY
EFpz뗥`o)k1W}RLYskuA8uf/K.<`ERhc
YZpKF]ޅI.Ɛ\h46r#DFGc+q6V[0elO>sʯfJ9Iwnӣ+O)v:Y3Ǯ܋OwVeFO^ruo+>NM{5Wr>kuzGz?f'vU_a9cvWey:c"7d1'Uօ=1 CajG;Kݏ~;_F=/R~0B3q(k"UTJӠQb2ez\
JӠQbֆ=s(+:C,f
WV 9@6s(\b}O6\8Wb6Jgh8Jc34+fA1%EԨ[I\`s;Ag949]s>wgЉk,نs]81 K `> 'g@XY
apae)\iSi+Ka>u#5"(Hz.փ~±H+MX&>FOxvmpeImI8zׁ;bp9գS,pp}MmM+M~G*ޱ +4߲Xx= +0vatZ]bÏYw\gլѻ Ș^[Ⱦ字c1wz bzEdUm=_/^`!iA@]=6C
ItxڝUkA6mHIE{I+~7z3`
zX!P(g/A=)zQM햸˛7ﷻ*K` ց#\(R_2B!*Ǡ%D1
M0?/<ėG÷
{j~WYU⍰@eƟ\YEo"?WW8
m'V/Le!lWM(@D@g&L>d_RJREvRw(ޟOy &ّ*Ԡӟ*Bya[h[h\GAQC=9ofzGJ)ؽFv)slH:hl=x0LL=\Lt4 uaK/Testing hypothesis about population parameters 00$v]jTesting Hypothesis about a Population Prevalence p w^A sample of n=100 adults is selected from Pakistan. In this sample 28 adults are hypertensive. Do the data provide sufficient evidence that the Government s figure is wrong, i.e., P>0.20? Test at 5% level of significance, that is, a=0.05. 4 Question:
Estimate prevalence==0.28
Hypothesized prevalence =0.20
Is the gap of 0.08= 0.280.20 considered statistically significant at 5% level?& u _Testing hypothesis about P}We need to calculate a test statistic
How many standard deviations have we deviated if the null hypothesis p=0.20 was true?
cMWhat is the likelihood of observing a Z=2.0 or more extreme if the Government s figure was correct?Elements of Testing hypothesistNull Hypothesis
Alternative hypothesis
Level of significance
Test statistics
Pvalue
Conclusion
Power of the test ut9Is there an association between Drinking and Lung Cancer?fWhat is the most appropriate and feasible study design in order to test the above research hypothesis?#Case Control Study of Smoking and Lung CancerQNull Hypothesis: There is no association between Smoking and Lung cancer, P1=P2
RRiSoYTest Statistic FA statistical yard stick which is computed based on the information contained in the sample under the assumption that the null hypothesis is true.
Knowledge about the sampling distribution of the test statistics is needed in determining the likelihood of observing extreme values for the test statistics in a given situation.% &!'"H=(#)$*%bL,Testing hypothesis about one population meanH0: m =16 vs Ha: m >16
Z= (sample mean hypothesized mean)
SE of the Mean
Under the null hypothesis and when n is large, (n>30), the distribution of Z is standard normal.
Pvalue
ConclusionZ?ZZtZ"t ` ̙33` ` ff3333f` 333MMM` f` f` 3>?" dd@,?" dd@ " @ ` n?" dd@ @@``PR @ ` `p>>c(
6 P
T Click to edit Master title style!
!
0
RClick to edit Master text styles
Second level
Third level
Fourth level
Fifth level!
S
0 ``
=*
0 `
?*
0D `
?*B
s*иh ? ̙33 Default Design0T4(
T
T
0 P
=*
T
0$
?*
T
6D `P
=*
T
6 `
?*H
T0h ? ̙33 (
l
Cd@
l
Cp
H
0h ? ̙33\@( B
6$ P
PInferential Statistics$
0
F Estimation: This includes point and interval estimation of certain characteristics in the population(s).
Testing Hypothesis about population parameter(s) based on the information contained in the sample(s).
&$ljH
0h ? ̙33g_P(
4
6 P
VImportant Statistical Terms $
0$ PP
Population: A set which includes all measurements of interest to the researcher
Sample: Any subset of the population
Parameter of interest: The characteristic of interest to the researcher in the population is called a parameter.
nR
J! kQH
0h ? ̙33`0(
6 P
REstimation of Parameters 8
0
lPoint Estimation
Interval Estimation (Confidence Intervals)
Bound on the error of estimation (???)
The width of a confidence interval is directly related to the bound on the error. 6^QH
0h ? ̙33B:p(
6 P
i9Factors influencing the Bound on the error of estimation ::
0
c!Narrow confidence intervals are preferred
As the sample size increases the bound on the error of estimation decreases.
As the confidence level increases the bound on the error of estimation increases.
You need to plan a sample size to achieve the desired level of error and confidence. &"*XH
0h ? ̙33
\Z(
\x
\ c$P
\
6 P
<$
\
0 0
4 OR or RR
Mean = m
Standard deviation = s
Difference between two population means
Proportion = p
Difference between two population proportions
Incidence H)H
\0h ? ̙33
H@(
r
S$P
&
6]
LSuppose the Government report that prevalence of hypertension among adults in Pakistan is at most 0.20 but you as a researcher believe that such prevalence is greater than 0.20
Now we want to formally test these hypothesis.
Null Hypothesis H0: P0.20
vs
Alternative Hypothesis Ha: P>0.206'  #H
0h ? ̙33$(
r
S`
r
S@P
H
0h ? ̙33d
(
r
SdP
r
S
x
<A}??Jh}H
0h ? ̙33
d0(
dx
d c$ 0
x
d
6P
" Pvalue= P[Z > 2.0] = 0.025
How does this pvalue as compared with a=0.05?
If pvalue < a, then reject the null hypothesis H0 in favor of the alternative hypothesis Ha.
In this situation we reject the Government s claim in favor of the alternative hypothesis.
$G( @`0Pp
@`
@` @H
d0h ? ̙33
(
l
CdP
l
Cd
H
0h ? ̙33t(
tl
t CDp
l
t C$ `
H
t0h ? ̙33F( &a}
l
CP
l
Ctp@
f
0" P
p
Alternative Hypothesis: There is some kind of association between Smoking and Lung cancer, P1P2.&c^H
0h ? ̙33~.&(


04
In the following contingency table estimate the proportion and odds of drinkers among those who develop Lung Cancer and those without the disease?x

<A[??$[H
0h ? ̙33l(
04:
^QUESTION: Is there a difference between the proportion of drinkers among cases and controls?
&_ Ul
0Ad??[# dH
0h ? ̙33
(
l
CtP
l
C
H
0h ? ̙33^0(
6T P
@Pvalue
0
XAn indicator which measures the likelihood of observing values as extreme as the one observed based on the sample information, assuming the null hypothesis is true.
Pvalue is also known as the observed level of significance.6>H
0h ? ̙33XP@(
6 P
@The level of significance (a ) 2!
0
Ta is known as the nominal level of significance.
If pvalue < a, then we reject the null hypothesis in favor of the alternative hypothesis.
Most of statistical packages give Pvalue in their computer output.
a needs to be predetermined. (Usually 5%) @/ME,H
0h ? ̙33=P}(
6! P
QType I and Type II errors
0T
&Type I error is committed when a true null hypothesis is rejected.
a is the probability of committing type I error.
Type II error is committed when a false null hypothesis is not rejected.
b is the probability of committing type II error.&u}H
0h ? ̙33p `(
x
<A
f??{yfH
0h ? ̙33ip(
64! P
?Power of a test
0!
d4The power of a test is the probability that a false null hypothesis is rejected.
Power = 1  b, where b is the probability of committing type II error.
More powerful tests are preferred. At the design stage one should identify the desired level of power in the given situation. H
0h ? ̙33{+#(
6 P
MFactors influencing the Power
0
h8The power of a test is influenced by the magnitude of the difference between the null hypothesis and the true parameter.
The power of a test could be improved by increasing the sample size.
The power of a test could be improved by increasing a. (this is a very artificial way) H
0h ? ̙33kc(
6 P
LMinimum Required Sample Size
0
MUsually a Sample size calculation formula is available for most of the well known study designs. Some software packages such as EpiInfo could also be utilized for the sample size calculation purpose.
It is extremely important to consult a biostatistician at the design phase to ensure adequate sample is considered for the study. NN$H
0h ? ̙33T
`(
`x
` c$DP
3
x
` c$3
\
`
S A??PH
`0h ? ̙33x=ml\Օfvq~$&ENC>L+mR8Ǔ=xQ6ۭ')ZZ!FPEjӂ
BUEV5ҊTCuV*5T+scf<
w3=w=>K2LZh&4Jh0tjXt>T.b=<1! "!.B1mT*K" .ClE\؆x
ĕW##v ڈkk1Dq⍈?qět
nDVMv<q3❈w!~n]&vCr8`,LY"ޜHwj
INoWnxa~<# ԁ!kIXp
>quu,`+^Q8a[~Ou%0 Uj/b٬0GZU{P:sxf(g6{3'FSc9ܿhH+ otoG_(vkXOJeN
MI3VڮFԌXô2R[,Ing#$J
CĢF$7\Q2r{uReT]5_i+>t.ylDyai"i7MA#&Jn2JY)YlByx7Uz5N~fbseF}MʺjBhhODǡqPq(GwFu1ZAo7sTh[@zurW5'A)T
V`Zu_خ5"CJRGpNVQ^fomi/i7vxޥY2gmu,nCY̗K/K0ϥF)Wm;//ǂ`dF]!#JteM#hr[h_?F%7E
N`<FEP!J3@/H0w23 ]fl2=8HыFojz<G);9ٹ
h,Ze4@:Rqrnr* pr$=Mڙ!{lbdA"s#(rQznM>p\^e;
~V`ij'=f9\oT+Ǔx97ÖUN+0%+6WQܪ]pvh9\H"V8[q$ZN+؏yV}KS;wNsOg(\m.xUܹU(.8zs{b5x"z9{5IqR\)fuZCꕮBKt[þXpP_v"wb)%J2X`gvvlkni3df&wn/neNd%3{^jk?O1?U$r;t%>SAKl^Do/)N%~pxm`J 6WIc>6M{t=`J]Pms>VTk8Z)]AzٻB]8ZR(}W
o>*%<,އzaYŝkwXs_qgrfΝ+ o^8yО:,U{hO#YXc+n sps g*hA i?}b?8^£Y!v"`G+8?0ev"`UG+mi"
lQlD(x op7xse2ܩ>efːgs~۹FmY
qA~ɹsbsO1o~ɹzyhȳ=yw5옘gKyըrSILSlS)&7_rnzy"Yըcފ7{Ճvc+iVەpN%Quγ
ʸkO;ԣVPgAwAL]OFoP/8UsRg/%+WܣB=\Ԁwq=ż?b5b1y/pW.Fwb&ż;RFƾ%/qe%i妠VReq~J%SjF<#{X/OkbkמG:o8?
VDDetpߞ˛7Z
ts _ᢗIw)AgA{ {ѿ SeTY/T[Pzt#.w}B=A3(Ց5_+ݒcCh2GX:ak:W˵&u{dv4ȣ]y1wnxn)}1
)#?cS˦EuB!B!B!B"_>6O7kaMs_k0˟cv'6ljছ>à3t߇=ҷ!<(D_֧"
;ZPO
*sIᔽ@Ҿo89Féx:7 R.h3MkVYH:7XB!B!B!$~x]LU{ۯPlcc0yp5\hnb4\H13{S7c 1>,M9;s=.gNG(A<(.عuWNC]8hzI A8R*lT7\tG7/4\= zS:
u( vKR߅'{ְ:߬Z@5RP*nC܀kz$3 7#nA܊X
DC2NX=0QVjGJpibLaxfToq}jr3p)IdSү)N9xQ4IiTZAqMYJqӒȪ$>eZ:)frs.Ly+G8h{?Ƚ
uWo!
FIZ+ӼR֍8jQI$9_~q;ҥlޓ<,re!LhY8?ۜwڜI6eg.o\[8 GYkbkk .O /
;2NrJ6!ϼ ogXq.#.q}ZO\e2{TtBUvM5ޗAlnz}9C=Ce7ivbd?E~=+n!D1vOPrQ
aF0Sx" <E݊5D1l/QܸDRD\`
X=]!١}9vPpnwl2up>înPѳzy'?4GFULΒvuw[D5ͰiA~ܭ֞<ꦴrgw$S?Q
;fS"톒*[5^ENY!` v7ڄxoӶml0tp΅n5~}ogvG&D4/yF=OP_`ͰlWG[i:
u$Qǘ,NN;~2Fȳ!}lKǢ4 ~GWM
G5ۛho [m
ҕdنxk/R>o
qiql/,}^}`cx3rH=ՙǠe:pw?AaʋEkh߬πQR8ii
jˮ@wEWːgqr;@[1tzTx\}l[?=N
i(MHKhRХMK]Ʃ];tXvLHhiMhںI4DX#1m,UAyNS4gwϽ{߽s.x%'n!r2l2DHS\.gs.>UTڐɋkns
*@*[ZtZ]Ձ.Vi׃@AaՠF
J5U*ǇnJE[NI4:]#g`4S~FFo\5z>c
Ql^6 M37S/E2T>@I7<5ZP0>ӡyۯ)\^Ҕj7w}6X4kt#54+dw}ҩLj0ޝJSǒYsN2x1AƊ_s/Zwczs`_:ƾVY=tVADnn]=4L?)KX4M]%ɷHYF'<$Gx#n%Qf_M[6%&Rߙ!&M)<.z+zb8Srcgi]
~#
vlbt¬?k0SY6oi
j7lvmFh;xYn
"'KbyyQ`
w!tE~t~E!ڂrMϪv68y0_JX\,{dj?xNAwR헓ҥ:a38SJ?wg9*..MȖQ55H&*'MHnO'g^DU\ Jѽf}_S:J.o5н{u<>n鱖cXLca.{c(,]=yXSCY3^S?rFuih7cPqxb<%*Yyʷ7Bv'&
h24V\̎'34}M1{.v^ӦhS3r#yʑ`==nٱa6d3RXb:<'jP܋VN:Z*l\nϞ++dw~幡_i擿p]d7R$cIU5?4AE_[4tDDe7nԲ Q])_ۦ77[Z6Z7CĂHyCSS8قk[ZXc` #+!e1OIH=XbܪdQs[u]x:יuTa?/mTZ
ad/2<悯Fuyp$$Ƿ#hPszdn
ʙq9%.E^גq#rW)bnYaqD>D(ꬊGDI1;.xc'f`Bڇ~8>`{] !,ZITܖxJp1ba/y;<}nXV3U1B3e]UμHy<chLQ!Ú#6P'[{lRN{1{=±3~ bylqq_ ߶u~]GṆۛ9'/nZ\p;J.1Cz9m?x8T:<&Gj{EjO`kя=Ki;ç'6]y
tǽ/F.0^9ɉovC?=]kܛ< Hp$8Csѿ0wr>Er
/S}Ͽ#FwģI3(ߣxu}K~1ISUDv(6څ.\p
.\jT9"xYoW?wf?۴NN]'M1RhXiKi챳tή,RHHj6
(< R_U'H2Ν;_7jFf=9{c_TWNN4%"< K~MCaq< 4`J[˝]ΒD&=NzTO;!DLA}0]{'.Fe߿;;~k()'b&wڟh&}m}9ǴW,g`w}~l7ֆ}m9s@tc^~
:HDAp/`#(p8<>0 8 <
 )~S4<ΰ׀s<<
x3s
Ӛ+EהbybiJFAկ}you!QjkjV~;d峥jpm9ctD*utAs]/{Ջ%sIs.Ϝo89wi{9[Xru׳Kx1ϛ9Eǻ$]3Wry\q5df %[v
7²^)^YҎW*J1?nts^sgnzR)"K%,w[PnE}AAy)R6w"j\baڹB$G7h6sNsWzvKnoTbgh@
cqQEY9BU[ES{G1ʠ˲^ioDH3*[F䮑u^qo:SFN2_I`߂;>`ko?1<崵r95*>A_lq@M'2r!M337dHEd)[K\]W<:zyƳgeLWw)QRsNa3Eʖ9EyZ
HJDM^e$k
uaBP}
R]_hqR4+h>2H`(gmoj%͜;'Xi
;Xn2>J]~!&pt>{s3kxo1dcǒt4ic`#?G\u9 +.뻾]:v*vrxht+6N>^O^uۻ3$3m'coh /"*cP=u N7DU[GL'Di\&((\HgcИɑҝ:dڸ9F9rM@mH88HUՉfΒݬV&*ᕯ(~2ftWI.}$"80?5㑃^>prR5Tf*m7Aj`?G.R&)K0L埠w8}QhW'J"鷸Გaioաkw/`1r0A) KWe%M!0!+.O%`Y6]aglpE@D2v<6+fVm[^ua09ZtFiRoUv =BzD
(
d
University1o
Times New RomanSymbolEuclid SymbolDefault DesignMicrosoft Word Document%Basic Elements of Testing HypothesisNo Slide TitleNo Slide TitleNo Slide TitleNo Slide Title0Testing hypothesis about population parameters :Testing Hypothesis about a Population Prevalence “p” A sample of n=100 adults is selected from Pakistan. In this sample 28 adults are hypertensive. Do the data provide sufficient evidence that the Government’s figure is wrong, i.e., P>0.20? Test at 5% level of significance, that is, =0.05. Testing hypothesis about Pg What is the likelihood of observing a Z=2.0 or more extreme if the Government’s figure was correct?Elements of Testing hypothesis:Is there an association between Drinking and Lung Cancer?.Case Control Study of Smoking and Lung CancerNo Slide TitleNo Slide TitleTest Statistic No Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleTesting hypothesis about one population meanFonts UsedDesign TemplateEmbedded OLE Servers
Slide Titles 6>
_PID_GUIDAN{109E05A6B71043C88579D17FAF2FC81F}Times New Roma_0Eugene .2
n5 Dr. M. H.s
)
!. .2
n Rahbar
.Times New RomanCwCw w0 ..2
Professor of Biostatistics
. ..2
De0VEquation Equation.DSMT40*MathType 5.0 Equation0[[Document Word.Document.80.Microsoft Word Document0d^MS Org Chart OrgPlusWOPX.402MS Organization Chart 2.00}zEquation Equation.DSMT40*MathType 5.0 Equation/0DTimes New Roman̵00DSymbolew Roman̵00 DEuclid Symboln̵00
@n?" dd@ @@``hlV
2$ET
j lSu$$$$2$5OD5gzu2$].
hQ*$$2$G0IdR;o
2$iA@]=c$ иf@Ug4!d!d0t
ppp@<4BdBd<4!d!d:2___PPT9/0?%O=$Basic Elements of Testing Hypothesis %$fDr. M. H. Rahbar
Professor of Biostatistics
Department of Epidemiology
Director, Data Coordinating Center
College of Human Medicine
Michigan State University* !"E<G>aK/Testing hypothesis about population parameters &0.(f$v]jTesting Hypothesis about a Population Prevalence p "64(fw^A sample of n=100 adults is selected from Pakistan. In this sample 28 adults are hypertensive. Do the data provide sufficient evidence that the Government s figure is wrong, i.e., P>0.20? Test at 5% level of significance, that is, a=0.05. D Question:
Estimate prevalence==0.28
Hypothesized prevalence =0.20
Is the gap of 0.08= 0.280.20 considered statistically significant at 5% level?4 u _Testing hypothesis about P"(f}We need to calculate a test statistic
How many standard deviations have we deviated if the null hypothesis p=0.20 was true?
~}cMWhat is the likelihood of observing a Z=2.0 or more extreme if the Government s figure was correct?*ec(fElements of Testing hypothesis ftNull Hypothesis
Alternative hypothesis
Level of significance
Test statistics
Pvalue
Conclusion
Power of the test &us9Is there an association between Drinking and Lung Cancer? :9ffWhat is the most appropriate and feasible study design in order to test the above research hypothesis?gf#Case Control Study of Smoking and Lung Cancer".(fQNull Hypothesis: There is no association between Smoking and Lung cancer, P1=P2
RRiSoYTest Statistic *fFA statistical yard stick which is computed based on the information contained in the sample under the assumption that the null hypothesis is true.
Knowledge about the sampling distribution of the test statistics is needed in determining the likelihood of observing extreme values for the test statistics in a given situation.GF% &!'"H=(#)$*%bL,Testing hypothesis about one population mean ,fH0: m =16 vs Ha: m >16
Z= (sample mean hypothesized mean)
SE of the Mean
Under the null hypothesis and when n is large, (n>30), the distribution of Z is standard normal.
Pvalue
Conclusiond?ZZtZ"t$(
r
S@
r
S@P
H
0h ? ̙33d
(
r
Sd`
r
SPp
x
<A}??i}H
0h ? ̙33
dD(
dx
d c$P 0
d
6@x
&" Pvalue= P[Z > 2.0] = 0.025
How does this pvalue as compared with a=0.05?
If pvalue < a, then reject the null hypothesis H0 in favor of the alternative hypothesis Ha.
In this situation we reject the Government s claim in favor of the alternative hypothesis.
8E ( @`0Pp
@`
@` @H
d0h ? ̙33
(
l
CdP
l
Cd
!"#$%&'()*+,./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_abcefghijklmnopqrstuvwz}~Root EntrydO),v@Pictures\Current UserH&SummaryInformation(
PowerPoint Document(1DocumentSummaryInformation8,d`0fFDocument Word.Document.80.Microsoft Word Document0VEquation Equation.DSMT40*MathType 5.0 Equation0[[Document Word.Document.80.Microsoft Word Document0d^MS Org Chart OrgPlusWOPX.402MS Organization Chart 2.00}zEquation Equation.DSMT40*MathType 5.0 Equation/0DTimes New Roman̵00DSymbolew Roman̵00 DEuclid Symboln̵00
@n?" dd@ @@``hlV
2$ET
j lSu$$$$2$5OD5gzu2$].
hQ*$$2$G0IdR;o
2$iA@]=c$ иf@Ug4!d!d0t
ppp@<4BdBd<4!d!d:2___PPT9/0?%O=W$Basic Elements of Testing Hypothesis %$fDr. M. H. Rahbar
Professor of Biostatistics
Department of Epidemiology
Director, Data Coordinating Center
College of Human Medicine
Michigan State University* !"E<G>aK/Testing hypothesis about population parameters &0.(f$v]jTesting Hypothesis about a Population Prevalence p "64(fw^A sample of n=100 adults is selected from Pakistan. In this sample 28 adults are hypertensive. Do the data provide sufficient evidence that the Government s figure is wrong, i.e., P>0.20? Test at 5% level of significance, that is, a=0.05. 4 Question:
Estimate prevalence==0.28
Hypothesized prevalence =0.20
Is the gap of 0.08= 0.280.20 considered statistically significant at 5% level?& u _Testing hypothesis about P}We need to calculate a test statistic
How many standard deviations have we deviated if the null hypothesis p=0.20 was true?
cMWhat is the likelihood of observing a Z=2.0 or more extreme if the Government s figure was correct?Elements of Testing hypothesistNull Hypothesis
Alternative hypothesis
Level of significance
Test statistics
Pvalue
Conclusion
Power of the test ut9Is there an association between Drinking and Lung Cancer?fWhat is the most appropriate and feasible study design in order to test the above research hypothesis?#Case Control Study of Smoking and Lung CancerQNull Hypothesis: There is no association between Smoking and Lung cancer, P1=P2
RRiSoYTest Statistic FA statistical yard stick which is computed based on the information contained in the sample under the assumption that the null hypothesis is true.
Knowledge about the sampling distribution of the test statistics is needed in determining the likelihood of observing extreme values for the test statistics in a given situation.% &!'"H=(#)$*%bL,Testing hypothesis about one population meanH0: m =16 vs Ha: m >16
Z= (sample mean hypothesized mean)
SE of the Mean
Under the null hypothesis and when n is large, (n>30), the distribution of Z is standard normal.
Pvalue
ConclusionZ?ZZtZ"tj@( B
6$ P
TInferential Statistics"(f
0
P Estimation: This includes point and interval estimation of certain characteristics in the population(s).
Testing Hypothesis about population parameter(s) based on the information contained in the sample(s).
0$ljH
0h ? ̙33yqP (
4
6
fImportant Statistical Terms .(f$
0$ P0
Population: A set which includes all measurements of interest to the researcher
Sample: Any subset of the population
Parameter of interest: The characteristic of interest to the researcher in the population is called a parameter.
pR
J! kQH
0h ? ̙33`H(
6 P
bEstimation of Parameters ,(f@
0 P
lPoint Estimation
Interval Estimation (Confidence Intervals)
Bound on the error of estimation (???)
The width of a confidence interval is directly related to the bound on the error. >^QH
0h ? ̙33h`p(
6 P
y9Factors influencing the Bound on the error of estimation $:8(f
0
y!Narrow confidence intervals are preferred
As the sample size increases the bound on the error of estimation decreases.
As the confidence level increases the bound on the error of estimation increases.
You need to plan a sample size to achieve the desired level of error and confidence. <"*TH
0h ? ̙33&
\f(
\x
\ c$`
\
6 P
<$
\
0 p0
2OR or RR
Mean = m
Standard deviation = s
Difference between two population means
Proportion = p
Difference between two population proportions
Incidence V)H
\0h ? ̙33
NF(
r
S$
,
6s
NSuppose the Government report that prevalence of hypertension among adults in Pakistan is at most 0.20 but you as a researcher believe that such prevalence is greater than 0.20
Now we want to formally test these hypothesis.
Null Hypothesis H0: P0.20
vs
Alternative Hypothesis Ha: P>0.20:( #H
0h ? ̙33r4! EGav^
(
d
0fFDocument Word.Document.80.Microsoft Word Document
!"#$%&'()*+./0123456789:;<=>?@ABCDEFGOh+'0
hp
@LX
dpxbFundamentals of Research Project Planning: Hypotheses, Questions, Objectives, and Indicators. sInformation Systems UnitCHSectEugenet108Microsoft PowerPointnit@Pg@=@.FR@pK"vTG$ oM :& &&#TNPP0D
&
TNPP &&TNPP
и !&G&ww@
oCwCw w0 &Gy& q`@ Times New RomanCwCw w0 f.2
Basic Elements of Testing'(1 ' . f.2
N
Hypothesis.! .I@h fTimes New RomanCwCw w0 .2
n5 Dr. M. H.s
)
!. .2
n Rahbar
.Times New RomanCwCw w0 ..2
Professor of Biostatistics
. ..2
Department of Epidemiology
. .:2
"Director, Data Coordinating Center
. .2
GCollege of Human Mediciney
#
. .2
}Michigan State Universityy#
."System
!$)Bw&TNPP &՜.+,D՜.+,D
zOnscreen ShowThe Aga Khan
H
0h ? ̙33t(
tl
t CD
l
t C$ `
H
t0h ? ̙33T( &a}
l
C
l
Ct 0
t
0"
0
Alternative Hypothesis: There is some kind of association between Smoking and Lung cancer, P1P2.4c^H
0h ? ̙33LD(
$

04
In the following contingency table estimate the proportion and odds of drinkers among those who develop Lung Cancer and those without the disease? x

<A[??0 0[H
0h ? ̙33t$(
04l
^QUESTION: Is there a difference between the proportion of drinkers among cases and controls?
._ $U $l
0Ad??``P0dH
0h ? ̙33
(
l
Ct0
l
C
H
0h ? ̙33p 0(
6T `0
NPvalue * f
0 @`@
\An indicator which measures the likelihood of observing values as extreme as the one observed based on the sample information, assuming the null hypothesis is true.
Pvalue is also known as the observed level of significance.:>H
0h ? ̙33@(
6 `0
@The level of significance (a ) :!ff
0 P
~a is known as the nominal level of significance.
If pvalue < a, then we reject the null hypothesis in favor of the alternative hypothesis.
Most of statistical packages give Pvalue in their computer output.
a needs to be predetermined. (Usually 5%) j"ME*H
0h ? ̙33GP(
6! `0
UType I and Type II errors f
0T 0@
,Type I error is committed when a true null hypothesis is rejected.
a is the probability of committing type I error.
Type II error is committed when a false null hypothesis is not rejected.
b is the probability of committing type II error.,u}H
0h ? ̙33p `(
x
<A
f??Kp&
fH
0h ? ̙333+p(
64! `
MPower of a test"f
0! Pp
p4The power of a test is the probability that a false null hypothesis is rejected.
Power = 1  b, where b is the probability of committing type II error.
More powerful tests are preferred. At the design stage one should identify the desired level of power in the given situation. H
0h ? ̙33E=(
6
[Factors influencing the Power"f
0
t8The power of a test is influenced by the magnitude of the difference between the null hypothesis and the true parameter.
The power of a test could be improved by increasing the sample size.
The power of a test could be improved by increasing a. (this is a very artificial way) H
0h ? ̙33}u
(
6
\Minimum Required Sample Size$(f
0 @
MUsually a Sample size calculation formula is available for most of the well known study designs. Some software packages such as EpiInfo could also be utilized for the sample size calculation purpose.
It is extremely important to consult a biostatistician at the design phase to ensure adequate sample is considered for the study. NN$H
0h ? ̙33T
`(
`x
` c$D
3
x
` c$3
\
`
S A?? qH
`0h ? ̙33rt "#%`z\#!%K*Hb .iownL
0