I want to concatenate two datasets called ANSUR one for females and the other for males. But before doing that I would like to know which statistical points I should consider being able to consider the concatenated sample as one set of samples. These datasets contain 93 anthropometric measurements such as weight, height, waist circumference, etc.
to know if all features have significant differences or not, which statistical tests I would apply. Thank you.
Soheila Ahmadi
Unpaired t test (when the data is parametric)
Mann–Whitney U test ((when the data is non-parametric)
Subhankar Banerjee Thank you so much for the response.
Most of these features do not have a normal distribution and homogeneity of variance.
Is it better to normalize them with natural log transformation and apply parametric tests or directly apply non-parametric models?
Would you also explain why an unpaired t-test is needed for that?
Soheila Ahmadi
Don't normalise your data with natural log transformation, because it will create some complexities...
You may use nonparametric test Mann Whitney U test.
Unpaired t test means group 1 and group 2 have unequal sample.
No of male participants= 17, no of female participants= 25
Hello, I suggest you first use the coefficient of variation, to verify that the data in both bases do not present high levels of variability. And after this use Unpaired t test, which is the test used in anthropometric studies.
all the anthropometric variables generally show sex differences. indices like BMI in some samples do not differ between sexes. Unpaired Student's t or analysis of variance can be used to test the difference
If the samples have equal variance then an independent-sample t-test (Student’s t-test) could be used.
If the samples have unequal variance then the Welch’s t-test can be used, as this does not assume identical variances.
If the samples are not normally distibuted, or if the other requirements of the above tests are not met, then the non-parametric Mann-Whitney U test should be used
With this software you can detect two mixed gauss distributions in data. An example (heights of male and female 16 year old people) is included.
www.lerenisplezant.be/fitting.htm
Hello!!!! Many of these variables do not meet the normal distribution assumption and neither do they meet the homogeneity of variance assumption. I suggest that you test whether these assumptions hold and if not you can use a non-metric test like Mann Whitney's U. If the aforementioned assumptions were met, you could use the Student's t-test for independent samples.
- Badges
- Science topic