I am working on diabetes patients' data having two groups. over a period of 3 years, even though the mean HbA1c at the beginning and at the end is more or less the same. But, one of the groups has more spikes in between, as I could see on the chart by Quarterly average HbA1c. I would like to get statistics to report. what would be the best statistics to confirm that one has more spikes in terms of numbers?
Thanks in advance
Hi,
I cannot verify from your data, but don't you already self-answer the question: distribution? Hence look into sd() or 2nd derivative of measurement instead of mean.
JW
There may be a stat, however usefull? might be but why not stick with the graphs and the old interocular shock test? If the difference is so great it hits you between the eyes,you do not need some silly stats test.
You might do better to superimpose or facet the data sets.
1 Like
Yes, always look before leaping. But we all suffer from lying eyes, so always good to pick a test to show if any visual differences are not illusory
1 Like
Thanks,
yes SD is higher for one set of data compared to the other.
for the 2nd derivative test in R, I browsed the net. I could not find any clue to calculate for a series of data, eg one column in a dataset having multiple rows
would be great if you have any suggestion
That depends on 1) what is meant by spike and 2) whether the data is continuous (5.8, 7.3, 6.0, 10.1, etc) or categorical (normal, high, ER, for example).
A spike could be defined as a value higher than both the preceding and following results. Or it could be a "new record." It's worth thinking about more abstractly, though.
f(x) = y, just like school algebra, is a useful construct. Here, x is the time series of readings, each of which is associated with a group and y is some measure of similarity. f is the function or functions to derive y from x.
Knowing y will allow an answer to the question *how similar are the two groups in terms of the HbA1c readings?" Given that the mean to start and end are "more or less the same", do they trace distinct paths to get there? (By the way, use a t.test to put a measure of the likelihood of the means being different. Is bumpiness the best measure?
For example, Group A may follow a path with lots of spikes but the amplitude is small, while Group B has fewer spikes but the swing is quite large.
This is likely a class of problem that has been addressed before. This review may be helpful.
1 Like
thank you very much for your support
1 Like
Thank you very much for taking the time in explaining the concept. I will work on this
thanks once again
1 Like
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.
If you have a query related to it or one of the replies, start a new topic and refer back with a link.