One Sample t Test (SPSS Exercises)
According to tradition in biostatistics, p-values (in spss it is significance (sig.)) which are
lower than 0.05 (α) are significant, meaning that we will likely still find differences if we
collected another sample of participants.
When using SPSS we are no longer confronted with a ‘critical value.’ Instead, we can
simply observe that the p-value is less than ‘.05.’
p ≤0,05 Ho rejected
p>0,05 Ho fail to rejected (accepted)
An internal physician took measurements of Fasting Blood Sugar (FBS), Cholesterol (cho),
Triglyceride (Tg) and Iron (I) in 723 inpatients. Can it be said that the biochemical
measurements of the doctor's patients differ from the expected values? Comment at 5%
significance level.
Expected
Measurement Values
FBS 90
Cho 200
Tg 150
I 85
Assumptions: Variables should be normally distributed to use one sample t test so firstly the
assumption should be checked.
FBS (n=521)
A doctor collected Fasting Blood Sugar serum analyse values from a random sample of 521
apparently healty subjects. The mean and standard deviation computed from the sample are
97.46 and 11.21 units/100 ml, respectively. They want to know whether they can conclude
that the mean of population from which the sample of serum amylase determinations came is
different from 90.
Ho:µ=90
H1: µ≠90
Fasting Blood glucose is normally distributed
so we used parametric test: one sample t
test.
T-Test
1
One-Sample Statistics
N Mean Std. Deviation Std. Error Mean
fasting blood glucose 521 97,4633 11,21441 ,49131
p (sig.) ≤0,05 Ho rejected
p (sig) >0,05 Ho fail to rejected (accepted)
Ho rejected. Statisticaly, fbg is 97.46 (11.21) is higher than expected (90) value
(t(520)=15.191; p<0.001).
Cho
A doctor collected Cholesterol serum analyse values from a random sample of 521 apparently
healty subjects. The mean and standard deviation computed from the sample are 179.69 and
35.73 units/100 ml, respectively. They want to know whether they can conclude that the mean
of population from which the sample of serum analyse determinations came is different from
200.
Ho: µ=200
H1: µ≠200
It is normally distributed.
T test
2
p<0,001
p (sig.) ≤0,05 Ho rejected
p (sig) >0,05 Ho fail to rejected (accepted)
Ho rejected. Statisticaly, cho values 179.69(35.73) is less than expected (200) value
(t(520)=-12.976;p<0,001).
Tg
A doctor collected Tg serum analyse values from a random sample of 499 apparently healty
subjects. The mean and standard deviation computed from the sample are 145.68 and 59.86
units/100 ml, respectively. They want to know whether they can conclude that the mean of
population from which the sample of serum analyse determinations came is different from
150.
Ho: µ=150
H1:µ≠150
Our sample size is very large (n>300) so we can accept tg values normall distributed. We
can use parametric test.
3
p (sig.) ≤0.05 Ho rejected
p (sig) >0.05 Ho fail to rejected (accepted)
p=0.108>0.05 Ho fail to rejected. Statisticaly, the tg values avearagely 145.68 (59.88) is
not different from the expected (150) value (t(498)=-1,610;p=0,108)
Iron
A doctor collected Iron serum analyse values from a random sample of 493 apparently healty
subjects. The mean and standard deviation computed from the sample are 86.72 and 29.61
units/100 ml, respectively. They want to know whether they can conclude that the mean of
population from which the sample of serum amylase determinations came is different from
85.
Ho: µ=85
H1: µ≠85
4
Our data is normally distributed.
p (sig.) ≤0,05 Ho rejected
p (sig) >0,05 Ho fail to rejected (accepted)
p=0,198 > 0,05 Ho fail to rejected. Statisticaly, the iron values avearagely 86.72 (29.61) is
not different from the expected (85) value (t(492)=1,288;p=0,198)
Ho: Treatment is not effective
5
H1: Treatment is effective
Hypothesis test (Sample)
Real Life (Population) Accept Ho (Treatment is Reject Ho
not effective ) (Treatment is effective)
Treatment is not effective Correct Decision False Decision-Type I
(Ho true) (1-α) error α
Treatment is effective False Decision-Type II Correct Decision (1-β)
(Ho false) error β
α: 0.05 usually we take (statistical analysis )
α: 0.01
α: 0.10 (small sample sizes)
α:0.001
β: related with our sample sizes
Before beginning to research we have to arrange β with calculation sample sizes.
β≤0.20
(1-β)≥0.80---- Power of your study (Power analysis)- It also means you use enough
sample size for your study. -