Question: How do you determine relationships, including quantitative relationships, between independent and dependent variables?

Neil Taylor answered on 17 Mar 2015:

Hi Filer2212!

Great question! I suppose that the method I prefer is to use graphs, as it allows a visual representation of the data that you are analysing! There are other methods such as “regression”, which is basically some maths to model the relationship… but I guess it’s all related!

An example:

Part of my current job is to look at a bunch of sensor data that has been sent in from a company for us to analyse. The data is from sensors called “accelerometers”, which gives us information about how fast something is accelerating in three dimensions (i.e. up/down, left/right, and forwards/back).

For this particular job, there are several accelerometers located along the length of a riser, which is a bit of pipe that goes from the seabed up to a ship on the top of the water. I’ve added a screenshot of what this looks like to my profile.

By using some maths and science (“fourier transformations”!) this data can be transformed into some more useful information – like what angle the riser is tilted at, and how far it has moved up/down, left/right, and forwards/back over time at the sensor locations! This means that we can work out what different locations on the riser have been up to over the past wee while (in this case, a few months).

By looking at the transformed data and plotting some graphs, we can work out several things. One of the things to look for is inclination-to-displacement ratio through time, which is basically saying “how much does the riser tilt for a given sideways movement”. The unit of this is “degrees per metre”.

If the ratio is increasing, then we are getting MORE tilt for a given movement. This would suggest that the soil supporting the riser is weakening over time.

If the ratio is decreasing, then we are getting LESS tilt for a given movement. This would suggest that the soil supporting the riser is strengthening over time – most likely because it is consolidating.

There’s also a whole lot of other stuff that is useful to measure here, including graphing inclination against displacement, before using linear regression to find out their relation to each other.

So that’s one example. Some other unrelated examples might include

– stress induced by bending
– expansion due to temperature
– friction force due to displacement

The way that I might get the data points in the first place to graph may be from:

– hand calculations
– spreadsheet software (Excel/MATHCAD)
– Finite Element Analysis software (FEA)

FEA software allows you draw an object, divide it into little blocks, and then ask the program to work out how the object will react to temperature/bending/stretching etc.

This is the sort of stuff I really enjoy doing! A bit of problem solving!

Hope this helps!