Consistency of Data

Put the data set into an excel spread sheet (or similar) and plot the data (I am assuming the data is in a format that can be imported easily).

Checking the first 2500 against the last 2500 is likely to give you no help whatsoever for all but the most simple processes being monitored.

Plot all the data, don't be stingy and try and shortcut the analysis otherwise the results will be worse than a random guess, you could be lending weight to the wrong conclusion. This is basic statistical and data analysis.

No!

How about comparing the first and the last value?

Check one in the middle too!

5000 samples collected at the same frequency is a nice start. After you plot these data points and visually examined for a trend, I would perform a Fast Fourier Transform analysis of the data. Remember, you will have to either add about 3000 null points or remove about 900 points to make this a Fast Transform. Adding null points is better because nulls will also be added in the analysis. This should help you to see if your data set has any periodic component. Next I would do a purely statistical analysis and see if the distribution resembles any of the known Probability Density Functions.

You must define the purpose of your analysis and choose a method of analysing your data that best suits the purpose....

http://www.excel-easy.com/data-analysis.html

http://www.qtiplot.com/doc/manual-en/c6433.html

I think he said that the samples were taken at the same frequency, so a Furier analysis won't apply here because it is intended to show the amplitude of each of the fundamental frequencies that compose a complex signal (waveform).

His best option is to feed his data into a graph in excel, he can establish trend and consistency from it.

You're correct that the OP made an ambiguous statement about the frequency of the samples taken. I assumed that they meant that the sampled data was taken at the same, repeated time interval (inverse of frequency). This is a mandatory requirement for performing a Fourier analysis. The highest frequency detectable would also have to comply with Nyquist's theorem. Only 5000 samples would also not produce a very fine frequency resolution at all in the analysis, too.

If instead the OP meant your understanding that the magnitude of a frequency was measured 5000 times then a Fourier analysis cannot be performed at all.

I take that back, one can always perform any analysis on any data set to produce a result. That result maybe useless or worse, misleading if you don't know the validity of the data or the limitations of the analysis technique.

It would be helpful to know what the 5000 data points represent and what you mean by consistent.

Why not plot the difference function over time domain? One can then perform FFT analysis on this to reveal periodic systematic errors in the data. If the FFT spectrum is low amplitude and mostly noise, then congratulations, there is essentially none of the periodic systematic errors. IF there is a constant offset error, no analysis can help you other than taking the median and the mean and comparing these to what the expectation value is. If the expectation value is 10.50001, but you have a mean of 5000 pts that is 10.60001, you still have a very precise, but inaccurate measurement. In such a case the error is (10.60001-10.50001)/10.50001 = 0.952%

You also must define (for yourself) what level of consistency is considered to be sufficient. If you want an accuracy into the ppm (parts per million) range, you must not only have a singularly precise instrument, it must also be found by measurements and calculations (not based on estimates, or approximations) to be accurate at that level prior to reporting any finalized data based upon the data set. You must incorporate a strict propagation of errors treatment all the way throughout the data analysis and final answer calculation. Otherwise, this will mostly be an exercise of a futile nature yielding nothing.

Central Tendency and Dispersion are kind of statistics whereas central tendency gets at the typical score on the variable and while dispersion gets at how much variety there is in the scores. Through this online course on Central Tendency & Dispersion using Data Analysis you will learn how to calculate central tendency & dispersion in practical using data analysis.
educba