> xzw5@ 4bjbj22 ZXX+HHH\DDDDLD\\2'''\~\~\~\~\~\~\$/^R`z\H'%'''\
\***'H\*'\**+Wl6HZXCrD_(X*\d\0\X`*`TZ\\`HZ''*'''''\\\\d*\\Sample size determination
R.Raveendran,
Chief Editor, IJP
Why is sample size important?
validity  too small a sample size will not yield valid results
accuracy only appropriate sample size will produce accurate results.
financial large sample costs more. It is a waste of money to test more subjects or animal than required.
resources it is not just money but other resources such as time and manpower are wasted by inappropriately small or large sample size.
ethics it is unethical to choose too small or too large a sample size. The validity of results obtained from a small sample size is questionable and hence the entire study will be a waste. Larger samples tend to waste resources and man power. The patients/animals are subjected to unnecessary hardship. Hence inappropriate sample size is definitely unethical.
What factors will affect the size of the sample?
degree of difference this is minimum important difference between groups which we want to detect. The difference between two means or proportions (control value and the expected test vaule).
type I error this is alpha error the chances of detecting a difference which does not exist (False Positive). It is usually set at 0.05 or 5%. This is also called the level of significance.
type II error  this is beta error the chances of NOT detecting a difference which it exists (False Nagative). It is usually fixed at 0.2 or 20%. 1beta is called power (10.2 = 0.8)
variation of results  Standard Deviations of control mean and the test.
drop out this rate has to be determined and appropriate allowance has to given.
noncompliance this will increase the variation and sample size has to be adjusted to the degree of noncompliance expected.
What methods can be used to determine the sample size?
arbitrary numbers not recommended
from previous studies may not be correct
nomograms & tables can be used; may not be flexible and accurate
formulas to be used. Varies with study design and analysis.
computer programs based on formulas; easy to use.
How to calculate the sample size?
Assess the difference expected
Find out the SD of variables
Set the level of significance (alpha)
Set the beta level
Select the appropriate formula
Calculate the sample size
Give allowances for dropouts and noncompliance Power calculation
Power is the probability that a study can reveal a difference between groups if the difference exists. Higher the chances of picking up the existing difference, more powerful is the study. Power is calculated by subtracting the beta error from 1. Power = 1 beta. Power is one of the elements required for calculation of the sample size. When a study is over, the power can be calculated (posteriori power calculation) to find out whether the study had enough power to pick up the difference if it existed. The minimum power a study should supposed to have is 0.8 (80%).
Posteriori power calculation can be carried out using computer programs. If a study reveals that the power is <80% and the difference between groups is not significant, the conclusion is that the difference between groups could not be detected rather than no difference between groups because the power of the study is too low to pick up the difference if it exists i.e. the probability of missing the difference is high and hence the study could have missed to detect.
Random Allocation
To allocate subjects or animals to different groups, randomization is followed so that each eligible individual in the population has the same chance of being allocated to a group. This eliminates selection bias. Haphazard allocation cannot be called random. Random number tables or computer programs can be used for random allocation. Simple randomization using computer generated random numbers are used commonly. Sometimes simple randomization may pose problems. For example, randomly allocated two groups may have a different male:female ratio. If it is important to have equal number of males and females in all groups, then one of the slightly complex randomization procedures such as block, stratified or cluster randomization should used.
Significance Testing
State Null hypothesis
Set alpha (level of significance)
Identify the variables to be analysed
Identify the groups to be compared
Choose a test
Calculate the test statistic
Find out the P value
Interpret the P value
Calculate the CI of the difference
Calculate Power if required
Null hypothesis (statistical hypothesis) states that there is no difference between groups compared. Alternative hypothesis or research hypothesis states that there is a difference between groups.
e.g. New drug X is an analgesic  (Research hypothesis)
New drug X is no better than a placebo (no difference
between the drug and placebo) (Null hypothesis)
Alpha is type 1 error and the acceptable limit is to be set. It is generally set at 0.05 and not above. If the P value is less than this limit then null hypothesis is rejected i.e. the difference between groups is not due to chance.
Think of statistical test as a judge. The accused (= drug under investigation) is not guilty (= null hypothesis) until the charges are proved. The judges decision depends on evidences (= data). He calculates the chances (P) of the evidence presented by the inspector (= researcher) being false (= probability of difference observed between groups being spurious). If this is <0.05, then he pronounces the accused guilty (the difference between groups is significant). If P>0.05, he awards the benefit of doubt to the accused and releases him (=the difference is not significant and the drug is thrown out).
Choosing an appropriate statistical test
There are umpteen number of statistical tests. Why cant we have just one that calculates the P value? The answer lies in the characteristics of data and type of analysis aimed at. There are different types of data (Continuous, Discrete, Rank, Score, Binomial) and the aim analyzing the data can also differ i.e. finding the significance of difference between means or medians or quantification of the association between the variables. The number of groups used in a study may vary and so does the study design (paired or unpaired). These variations make it difficult to have a single common test. The significance test has to be chosen carefully. Tests which are not appropriate for a given situation will lead to invalid conclusions. For example you cannot use a thermometer to measure the weight of a patient. Similarly a meter scale can measure the height of the patient but not the temperature. One can measure grains using a litre can as well as a balance. But which one do you think will be more precise?
Before choosing a statistical test, one has to work out the following details. This will help you choosing a test from the table given:
Determine :
Aim of the study
Parameter to be analysed 
Data type  [Continuous, Discrete, Rank, Score, Binomial]
Analysis type  [Comparison of means, Quantify association, Regression analysis]
No. of groups to be analysed  No. of data sets to be analysed 
Distribution of data  [normal or nonnormal]
Design  [paired or unpaired]
With the above information, one can decide the suitable test using the table given.
Significance tests can be divided into Parametric and Nonparametric tests. The former includes t test, ANOVA, linear regression and Pearson correlation coefficient whereas the later includes Wilcoxan, Mann Whitney U, KruskalWallis ANOVA, Friedman ANOVA and Spearman rank correlation. Those variables which follow normal distribution can be subjected to parametric tests and those which do not are suitable for nonparametric test. If the aim is to find out the association between variable, correlation or regression tests should be chosen; the difference between means or medians can be found out using other tests. If more than two groups/means are compared ANOVA should be used. These models/tests have variants also. For example there are many more ANOVA models (oneway, twoway, repeated measure, factorial etc) and one has to choose the appropriate one (remember the analogy litre can and balance to measure grains?)
Significance test to be used must be decided at the beginning of the study.
A study may need more than one test depending on the number and characteristics of the parameters studied.
Always spend enough time and brain to choose a right test
Inappropriate test will lead to invalid conclusions.
Calculating and Interpreting P
When the data are subjected to significance testing, the resulting value is called statistic. This can be t (t test), chi (chi square), F (ANOVA) etc depending on the test used. This statistic is used to find out the P value available from tables (statistics software can automatically calculate the P value). If the P value is less than the cut off value (level of significance i.e alpha error), it is considered that the difference between the groups is statistically significant. When P is <0.05, it indicates that the probability of obtaining the difference (between groups) purely by chance (i.e. when there is no difference) is less than 5%.
If P>0.05, the difference is considered statistically nonsignificant and it is concluded that there is no difference between the groups or the difference is not detected.
Nonsignificant result can be due to two reasons :
There is really no difference between the groups.
The study is not powerful enough to detect the difference.
Hence, one should calculate the power to conclude whether there is no difference or the power is inadequate. If the power is inadequate (<80%), the conclusion is the study did not detect the difference rather than there is no difference between groups
CI of the difference
Confidence interval of the difference (between groups) can be calculated. CI indicates the variation of the data. A wide interval implies a larger variation. If the CI includes 0 (e.g. 2.3 to 5.8), then it is unlikely that the difference is statistically significant (P>0.05). Many journals now insist on mentioning the CIs for important results.
Degrees of Freedom (df)
It denotes the number of samples that a researcher has the freedom to choose. It is based on a concept that one could not have exercised his/her freedom to select all the samples.
The concept can be explained by an analogy :
X + Y = 10 .(1)
In the above equation you have freedom to choose a value for X or Y but not both because when you choose one, the other is fixed. If you choose 8 for X, then Y has to be 2. So the degree of freedom here is 1.
X+ Y+Z = 15 ..(2)
In the formula (2), one can choose values for two variables but not all. You have freedom to choose 8 for X and 2 for Y. If so, then Z is fixed. So the df is 2.
df is calculated by subtracting 1 from the size of each group. For example df for Students t test is calculated by n 1 (paired design) and N1+N2 2 (for unpaired) [N1 size of group 1, N2 size of group 2]. The methods of df calculation may vary with the test used.
.
<=>\]eiQ R S [ ] p
IJPW\cgkzʿzpzpzpzh:hz56h:hR@56h.7hzCJaJh hgghgg h)5h8i#h h)h 5h)h0D5h)hz5h.7h:CJaJh:h0D56h.7h.75CJaJ h0D5h.7h.75h.7h.7CJaJh0Dhz++=>\]R IJ{@
^
&Fgd.
&FgdR@
&Fgd
&Fgd0Dh^hgd:
&Fgd:$a$gd.7gd.734z{?@LNe
,
<
?
@
]
_
e
f
g
]^_ablnױמממĎh)h5h)h.5hv hR@6hR@hR@56hR@hR@h
PhR@hR@6hzhhhh.hR@h)hR@5h)hz5h.7h:56CJaJhz566(+:PSex{0=>?@AG~ytlylh5h5>* h">*#jh5h5>*UmHnHu h.<>*h.7h.<5CJaJh.<h55h:ha556ha556h.7hz5CJaJhthAy5ht ht5hthz5hz56h:hz56h:h[56h.7hzCJaJ$;f>?@`a~ $Ifgd5h^hgd.<
&F
&Fgdt
&FgdK+,>?Tr}>~ɺxxpxllg_Xlh`h`hh`6 h`6h`hh6hhx(h*CJ aJ h*CJ aJ h55hxh.<>*hxhx>* hx>*jhx>*UmHnHu h5>* hi>*#jh5h5>*UmHnHuh5h5>*#jh5h">*UmHnHu ,?@Z[mnZ[zzzrjj$a$gd@<$a$gd@<$a$gd.7$a$gd*CJ aJ h,UhD5>*CJ aJ hDOJQJhpOJQJh@<h@<OJQJhOJQJh@<OJQJh@<h@<h*CJ aJ h@<h@<5>*CJ aJ h*hD5>*hDhD?hk5#jhD?hd5UmHnHuhD?hD5<cabcd~~~~yyywgdD$a$gdD?$a$gdD?pkdR$$IflJ
t
6 0644
lae4
""###7#W###Q$$$$$(((X)))))t,gdL$
&Fa$gd%:$a$gd%:$a$gdJi8$a$gd0y!#!'!k!l!n!!!!""""###$*$$$$$Z%%%%%%%%4&&&l((())))))ҿ{pihk5>*hkhkCJ aJ hL5>*h^5>*h%:5>*h#Y5>*ha5>*hJi85>*hM5>*h0y5>*h h\
8h\
85h\
8h 5hn85CJOJQJhz5CJOJQJhzhn8hc~hc~hc~h0yhUZ#h=4h.))++3+L+c+s+t+v+++++++C,D,r,s,t,x,!"h..........'/)/M/N/O/P/_/f/>0?0@0ȨȨșyh85>*hkA5>*hg5>*hVhVCJ aJ hV5>*hZx56>*hZxhZx56>*h>5>*h^5>*h5>*hZx5>*ha5>*hCE5>*h
P5>*h%/5>*h5>*hQ5>*hd5>*+t,u,"#VW....>0?0@0X0Y011=1>1W1X1)2$^`a$gd{cgd8gdV$
&Fa$gdZx$a$gdJi8@0R0W0X0000001;1<1I1P1V1o11112&2(262>2@2G2O2U2t2~2222222222Y3Z3333333444漵h~8h)h
5>*hO75>*hD5>*h'5>*h8Co5>*h{c5>*h.65>*h'5>*h85>*h8CJ aJ h8h8CJ aJ 1)2*2>2A222333333334444gdO7$^`a$gd{c$a$gdJi8/ =!"#$%P$$If
!vh5
#v
:Vl
t65
a
Y$$If!vh5
#v
:Vl
t
6 65
e48@8Normal_HmH sH tH D@D Heading 1$$@&a$
5>*CJ DA@DDefault Paragraph FontVi@VTable Normal :V44
la(k@(No List2>@2Title$a$5>*>B@> Body Text5CJ(OJQJ8Z@8@<
Plain TextOJQJj@#j)
Table Grid7:V0,,Z+=>\]RIJ{@^;f>?@`a~,  ? @ Z
[
m
n
Z[pq (*?AWY}~cabcd7WQ X!!!!!t$u$"%#%V%W%%%%&&&&>(?(@(X(Y())=)>)W)X))***>*A***++++++++,,,00p000p 00 0 0 0 0 00 00 0 0 0 0 0 00 00 0 0 0 0 00 00p000000000000000 000 00000p000000000000000000000000 0 0000000000000d0d0d0d0dp0d0d0d0d0d0d0d0d0d0d0d0d 0d 0dp 0dp 0d0d0d 0d 0dp0d 0dp0d0dp 0d 0d0C 0C 0Cp0C 0C 0 0p0C0(0p00(0000000 0 000 0p0 @0C0p0p0000 ,O900yyz?6L3 0wz
!)@04!#%()+[t,)24 "$&'*,423p0e0e
A5% 8c8c
?1 d0u0@Ty2 NP'p<'pA)BCDE"0e@ @ABC DEEFGHIJK5%LMNOPQRSTUWYZ[ \]^_ `abN E5% N E5% N F
5%
!"?N@ABC DEFFGHIJK5%LMNOPQRSTUWYZ[ \]^_ `ab@X2(
ZB
$
s*DZB
%
s*DZB
&
s*D ZB
'
s*D
ZB
(
s*D
ZB
)
s*D ZB
*
s*DZB
+
s*D
ZB
,
s*DZB

s*DZB
.
s*D ZB
/
s*DZB
0
s*DZB
1
s*DZB
2
s*DB
S ?@~ (?WY,99kt.99t/9M9t099t15<5t25P5(t$t%t&t'tt(t)V.t*t+Bt,t(6;owLTFNd#g#(!(T(V(****/+1++++,R[2=@D
^l _gcffmL%U%T(V(2);)L)N)8*;*S*Y*****y+}++++++,3333333333333333333333333333333333=+ + I)P)****6*>*F*G*O*U*t*~*****++++,+,Dr.RavindranDr.RavindranDr.RavindranDr.RavindranDr.RavindranDr.RavindranDr.RavindranDr.Ravindran
Dr. Ravindran Sleep Lab),2
T&W$>azYp.8dQ Qt J\ΰZ048{:(Tb!DS(4L$>6jL8dU#RXHSUȈy@f^`>o(.hWW^W`OJQJo(hH''^'`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHogg^g`OJQJo(hH77^7`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hHWW^W`OJPJQJ^Jo(''^'`OJQJ^Jo(hHo^`OJQJo(hH^`OJQJo(hH^`OJQJ^Jo(hHogg^g`OJQJo(hH77^7`OJQJo(hH^`OJQJ^Jo(hHo^`OJQJo(hH^`o(.
^`hH.
pLp^p`LhH.
@@^@`hH.
^`hH.
L^`LhH.
^`hH.
^`hH.
PLP^P`LhH.2
y@f QdU#RnAmm0SU6jL4azY0)WS(4L{:0&Z0b!D Jl " D \ ]\8%:x."O7L kD UZ#8i#Fz&x(5a5\
8wU8Ji8n8~8.<CEM,Uv_{cd@<Q`.6$ginD?:^d.7zK tDv)f6._9V'$lgg
PkA=?,  m
[pq,!(!(""@++++,P@UnknownGz Times New Roman5Symbol3&z Arial?5 z Courier New;Wingdings"qhC#ySHm%Om%O"24d++3QH(?AyElimination of biasDr.Ravindran Sleep LabT
Oh+'0p
,8
DPX`hElimination of biaslim
Dr.Ravindranf br.RNormal.dotn
Sleep Labn83eMicrosoft Word 10.0@Xu@rBKƖ@<"rm%՜.+,0hp
JIPMERO+A
Elimination of biasTitle
!"#$%&'()*+,/012345789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefhijklmnpqrstuvyRoot Entry FࢤCr{Data
.1Table6OaWordDocumentZSummaryInformation(gDocumentSummaryInformation8oCompObjj
FMicrosoft Word Document
MSWordDocWord.Document.89q