AboutAsmaa Abd Elhameed Ahmed Expertise Medical Statistics,how to choose your statistical test,
statistical tests using SPSS program
,Research design
,sample size calculation
Experience Demonstrator in Biomedical informatics &Medical Statistics department
,Alexandria university
Organizations Medical Research Institute,Alexandria university
Egypt
Education/Credentials B.Sc. of Medicine, Faculty of Medicine,Alexandria university
candidate for Master degree in Medical Statistics
Question QUESTION: Dear sir, we studied the effects of a drug given by four different methods in patients of asthma and measured their FEV1, PEFR, FVC and FEV1/FVC before and 1 hr after the drug. We have to compare FEV1, PEFR, FVC and FEV1/FVC after giving the drugs in the 4 groups and also the percentage improvement [(final value-initial value)*100/initial value]. We can use SPSS. Please advice which statistical tests should be used. Is ANOVA suitable for all these parameters? Thanks a lot.
ANSWER: Hi Dr Tariq,
yes, you will use one way ANOVA test if normally distributed data, if the data is not normally distributed we will use Kruskal Wallis test.
This for the percentage improvement and for the final effect also.
If you want to test the difference between before and after the drug in each group separately
we will use paired-t test.
---------- FOLLOW-UP ----------
QUESTION: Thanks for your quick response. Our statistician said that since FEV1/FVC is a ratio, ANOVA will not be a good test and we would have to use some other test. Please clarify.
ANSWER: hi,
in case of FEV1/FVC ratio it will be mostly not normally distributed,so you will use
"kruskal wallis test"
but check first by K-S test
---------- FOLLOW-UP ----------
QUESTION: Thanks for your expert advice. What is K-S test, what is it used for and how is it applied using SPSS.
Answer hi,
K-S test (kolmogrove-Smirnov test)is a test of normality
It is used to explore the distribution of data(if it is normally distributed or not)
To do it on SPSS:
analyze--descriptive---explore---plots--normality test with plots
In the output we look at p(significance) of k-s test:
if P<0.05 the data is not normally distrbuted and we run non parametric tests(e.g kruskal wallis test here)
If p>0.05 the data is normally distributed and we run parametric tests(e.gone way ANOVA test here).
