I have data of a right and left feet of few individuals which I have calculated manually as well as using a software program. List of data is given below. Can you please suggest which statistical analysis should I use to prove my calculated data is correct and similar with respect to actual measured data
Data I have:
1. For every individuals- Left and Right foot data
2. Data of every foot- (a)Measured Length (b)Measured Width (a)Calculated length (d) Calculated Width.
Is there no-one local to you who can help you with this? It seems to me that you are over-reliant on advice from ResearchGate, and, as you can see, much of it is well-intentioned but either misunderstands the nature of your research question or, worse, misunderstands statistical testing.
Please try and link up with your local statistics person. There is no substitute for discussing your study over a coffee at least a phone call.
Hello Parth Shah.
I think the best way to analyze your data is by, first, very if your variables are normally or non-normally destributed. Then you should apply a two dependent samples test - if variables are normally destributed, the two dependent samples t test - if variables are non-normally destributed you can use the Wilcoxon test (for a) versus c) and b) versus d)).
I hope I was helpful.
My data is normally distributed. One more doubt should i go for chi square distribution as one of my friend suggested it? or how should I proceed?
Thanks a lot for your reply ma'am.
The chi-square test is used for categorical variables. Since your variables are continuous with a normal distribution, I advise you to use the two dependent samples t test. You could categorize your data and use the chi-square but you will introduce bias in your analysis because information of your data will be lost by categorizing it...
A t-test for paired data (left and right) or a mixed model controlling for a radom individual factor could make sense to me...
I agree with Rita, go with a t test if they are normally distributed. BUT you need to do multiple t-tests ie one for left foot measured vs calculated length, one for right foot measured vs calculated length etc. Mixed model would work but for what you seem to want to do I'd just keep it simple with 4 t-tests.
Hi Parth,
It depends on your research question. If you are simply comparing each of your variables between the left and right foot, then a paired t test is what I would also suggest. Otherwise, if you intend to do any modelling, then use GEE (generalized estimating equations) or mixed effects model, to account for the correlation of the right and left foot in the same person.
Hi Parth
The previous people who have answered are partially correct. In fact, what you are trying to do is a method comparison study, ie measured versus calculated. This is primarily based on a Bland-Altman plot. Part of the Bland-Altman analysis is a paired sample t-test for zero bias, along with a check of the limits of agreement for clinical significance of the bias.
Adrian
Thanks Adrian, finally a good answer ;)
A test comparing measured and calculated values makes no sense.
You should instead estimate if there is a bias (possible the bias is size dependent! -you will see this in a Bland-Altman plot) and what difference between claculated and measured result one can expect (that's quantified by a probability model; this could possibly be a t-distribution, but this is hard to say without seeing the data).
Is there no-one local to you who can help you with this? It seems to me that you are over-reliant on advice from ResearchGate, and, as you can see, much of it is well-intentioned but either misunderstands the nature of your research question or, worse, misunderstands statistical testing.
Please try and link up with your local statistics person. There is no substitute for discussing your study over a coffee at least a phone call.
Basically you are comparing two methods of measurement. As suggested by others in this thread, an approach based on the analysis of residuals as described by Bland and Altman is appropriate. By the way: their paper is very well readable: Bland JM and Altman DG (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet (February 8): 307-310.
An alternative method, based on the analysis of a linear regression, has been described by Passing and Bablok in, again, a very well readable paper: Passing H and Bablok W (1983) A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in Clinical Chemistry, Part I. J. Clin. Chem. Clin. Biochem. 21: 709-720.
Both methods are quite intuitive, and the basic mechanics behind it easily grasped. Have a go and read these papers!
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis