Our secret blogger believes that better training in the use of statistics would help engineering graduates deal more effecitlvely with the uncertainties of the real world.

A little while ago I filled in aquestionnaire that included the question, “did your academic experience prepare you adequately for the working world?” Now it is a long time since I was at university, but my answer was yes, albeit with the caveat that I think it is really the employer’s job to do this training. I do think, however, that there was at least one area that my university course did not cover in sufficient detail.

The sort of people who gravitate towards engineering, are by definition good at maths and physics. Those who go on to engineering tend to do maths with mechanics at ‘A’ level and so don’t do statistics. These subjects are deterministic. 2+3 always equals 5. However when you add part A to part B, you don’t always get the same outcome.

Understanding, and coping with variation is a key fundamental for successfully engineered designs. This is something I’ve observed time and time again, from problems with assembling flat packed furniture, to resolving issues with a complex lever mechanism that had to allow precision movement but connected two floating structures.

The key to resolving most of the faults I’ve come across in my career has been an understanding of the tolerances and variability. When it comes to statistical analysis, there are no right answers and from experience that doesn’t sit well with the way we train our engineers. After I had been taught about the use of statistics to manage variation, and then started to spread the ethos of “on target, with minimum variation,” I looked back at my university notes and realised that we had, indeed, covered some of the concepts on a brief course wrapped up with a few other things. The content had been miniscule and the link to real world had been almost non existent. In short the module was almost useless as I couldn’t even remember I’d done it.

I’ve had a look at some current course syllabus summaries, and it looks like there may be some statistics getting taught, but from what I’ve seen of the graduates coming through (who are usually very good incidentally), the link to the real world importance of these things is still missing. So, I still think that in general employers should train graduates in workplace skills. However, I think that training in statistical methods to allow an understanding of how to deal with the variation we encounter in the real world, along with practical examples, is an area that could be covered more fully during our academic courses.

Furthermore, I think that given sufficient emphasis this could significantly improve the ability of our graduates to come up with capable designs.

I agree, the lack of statistical knowledge and practice is a weakness for engineers.

Off topic, but in light of the recent discussions about women in engineering, shouldn’t the secret engineer’s profile be 50% of the time a woman? To make it consistent with the probability of choosing at random from the overall population. Or 30% if the choice is from the engineers population.

Unfortunately, most Engineering departments are required by the University ordinances(turf wars) -to use the mathematics department to teach any module which has a mathematics content: and statistics invariably falls within this. The consequence is ‘”Well, they are only Engineers, we can use the worst lecturer we’ve got.” No matter how much statistics is in the syllabus, if the lecturer is NBG the students will suffer.

There are lies, damn lies and statistics too often in my experience sums up the reality of both teaching and application of statistics in many Engineering situations.

The most important/real question in most applications of statistics is simple. ‘How far from the ‘norm’ is a particular reading: and how far can we allow it to be before we have to do something about it?

In the analysis of staple fibre within a bale (in which there are about a billion individual fibres) it was common practice to extract a handful (I kid you not, that was the instruction for operatives), select a finger pinch from this handful, isolate 10 individual fibres from this and mmeasure the physical properties of these. A bale was approved or downgraded on this basis. Absolute nonsense!

Surely tolerances and ‘fits’ are best defined (in the same way as we eventually assessed fibre: will it card (the next process in the textile chain)

If it does, its good, if it doesn’t it isn’t!

If it fits its good, if it doesn’t it isn’t.

Reply to the editor: I am referring to the picture, which is a always the same and that of a man. When the blogger does not specify gender, like in this particular one, it would be fair and encouraging to see a woman’s profile, at least 30% of the time.

Agree with this – as a measurement and instrumentation person I find too many people quoting numbers that they have measured without any thought to the uncertainty of that data. These numbers are then taken as fact and used in different ways and can lead to misleading answers. I would advise anyone studyin any scientific or engineering subject to be conversant with statistics – particularly uncertainties, sampling, distribuiotns and so on.

I agree totally with Andy. In fact my immediate senior, and his, and indeed his in the bale of fibre episode (in the USA) were chemical engineers. I can recall the row with my immediate boss (which incidentally precipitated my taking an early bath from that firm!) about his interpretation of data. I pointed out the major weakness in his argument at a meeting of all our ‘section’ and then suggested that if he was right and I was wrong, then so had been Sir Isaac Newton: and that somehow the entire course of Western Technology had managed to struggle by without my boss’s contribution for over 300 years. he never forgave me! I could forgive him for being incompetent, as he was, but NOT for attempting to correct the spelling and grammar in my monthly report.

Hey Ho

Mike B

I’ve given my details, but I’m posting anaonymously to protect the identities of my children, who are both cited here.

I am clearly of a different generation.

I took a Statistics ‘O’ Grade at school, along with Arithmetic and Mathematics, when all three were separate subjects. Then I did statistics as part of my HNC Engineering course, then again as part of my undegraduate degree [admittedly, both taught by Maths lecturers], and finally, as part of a postgrad MEng course.

As a civil engineer, I used statistical analysis to check the validity of things like ground survey closure errors and micrometer readings of the thickness of protective coatings on structural steelwork. Are things like this no longer done?

And, on the bale example, I remember well doing AOQL graphs [Average Outgoing Quality Limit] to check the performance of components against upper and lower tolerance limits. Again, if this is no longer done by sampling and statistical analysis, then how is it done?

On the other hand, I know that my son went through school with straight As in Maths at all levels without doing statistics or forecasting, took a basic crammer in his undergraduate degree [Computer Science] and is now doing a Masters in Big Data a significant part of which is statistical analysis. At a recent parent/teacher meeting at my daughter’s scholl, I was almost derided by her maths teacher for asking why there was no statistics module in the course. The polite answer was that the topic was not valued very highly!