Statistics 5

[EireChch]

Interested to know what people’s take on statistics is?

Do many engineers use it on a day to day basis?

Does it reduce the need for engineering judgement? An older engineer has told me that while engineering judgement is important, it is becoming less acceptable these days.

I have used it on a few projects where we had lot of data. We looked at mean, standard deviation etc and worked out a value 95% confidence value which means that there is a 5% chance that you will have a parameter lower (or higher, depending on parameter) than your 95% confidence value.

The use of statistics is promoted in Eurocode but there is still room for an engineer to adopt a “cautious estimate” (i.e use judgement).

Has anyone had any bad experiences with statistics ? Was it killing your design?

 

[SlideRuleEra]

Does it [statistics] reduce the need for engineering judgement? An older engineer has told me that while engineering judgement is important, it is becoming less acceptable these days.

At least I meet the "older engineer" criteria, so here are my thoughts. Engineering judgement remains very important but is always secondary to engineering first principles and mathematics.

To me, the very best use of engineering judgement is to first perform the best analysis of a problem or situation possible (even when this means making reasonable simplifying assumptions).

Then apply engineering judgement to those results. This is light years better than only "thinking" about the problem then going with "gut feel" or some other such nonsense.

The concept of statistics enters the above picture by considering my own "simplifying assumptions"... Just how accurate are my simplifying assumptions? Can I expect my (approximate) answer to have first order accuracy? Or better? Or worse? At this point, engineering judgement kicks in.

An example of the type situation I encountered might involve an existing geotech report for a particular structure at a certain site. However, now we need to design/construct/use a very different type structure a distance away. Part of my job is decide if the available geotech report is reasonably useful, or not, what (if anything) to do about it in either case, then proceed.

This may not be exactly what you are asking, but an understanding how statistics represent real-world uncertainty is crucial.

http://www.SlideRuleEra.net

 

[r13]

I don't think statistic in daily design, but the design itself is statistic based - the load, the load factors, strength reduction and safety factors... Probabilistic study is very important is every steps of engineering, and it is a tool of making engineering judgement.

 

[Ron]

My practice is forensics and we get blasted routinely in litigation about not using random sampling (statistical) techniques for selection of sampling locations (but so far we have not lost our argument!). We select sample locations based on engineering judgment and experience. We know what we're looking for. If we used a randomizing technique, we would miss the problems.

While statistical analysis certainly has its applications, it should not substitute for engineering judgment and experience, and certainly should not be used to overrule first principles and proper engineering analysis.

 

[EireChch]

Sorry for the delay in response and I appreciate your responses.

SRE - I think that is a very good response and I believe it is how the steps of analysis should be performed. Principals followed by gut feel. I suppose one thing with judgement is that you get a better understanding on when you can rely a little more on judgement and when you cant. But then again, you fall into the risk of over relying

I use to work with a guy, and while he was a very good engineer, I think he relied on engineering judgement too much. Part of the problem is that he was running a team that was over worked and under resourced. We were always stretched in the time we had to complete reports and analysis and he would instruct me to do very simple analysis with his simplistic assumptions. I always thought we could have spent double the time on designs/assessments but he just flew through the designs. In saying that, I dont think there was every any issues with the work we done while I was there so he was obviously very good at it (or got away with it!)!

I recently got slated by an external review of a report for the below SPT graph. The reviewer was adamant that I had to use a "statistically derived characteristic design line". I havent actually got around to preparing a "statistically derived characteristic design line" for the chart. However, I believe what I had done was quite reasonable.

But my response was, "why do I have to use a a "statistically derived characteristic design line" and the response was that "any time you are assessing data you need to use statistics". The theme continued through the report review, with everything requiring assessment by statistics.

I found it quite weird tbh, that someone was so adamant that statistics had to be used.

 

[geotechguy1]

Suppose I have a commercial site that's 200mx150m in plan. I do 10 boreholes with a diameter of 150mm and carry out SPT's at 1.5m intervals. Lets say, rock is at 10m.5m, so I get 7 spts a borehole. There is 300 thousand meter's cubed of soil above the rock. My boreholes allow me to see less than 2 meters cubed of soil which is less than 0.0006% of the soil. The SPTs have tested even less than that.

What use are statistics?

I suppose it could help but I've never found any use in doing anything other than averaging

 

[GeoPaveTraffic]

I would ask the reviewer what type of statistics he/she would like to see used? Should it be an average, a weighted average, a 1/3 and 2/3's split of the data, etc.?

Statistics can be used in some situations, but those situations are generally limited in the real world of practicing geotechnical engineers.

The reviewer's comments remind me of the saying: Measure with a micrometer, cut with a chainsaw.

Mike Lambert

 

[SlideRuleEra]

ErieChch - Thank you. Perhaps what the reviewer means by a "statistically derived characteristic design line" is a best-fit, mathematically calculated linear regression line, similar to the green line that I "eyeballed" on your graph:

geotechguy1 - For your example, if all I had was the 70 data points (10 borings with 7 STPs in each),for my own use, I would prepare a graph that presents all the info graphically, on one page. Here is a sample, with data I made up:

The height of each vertical line of the graph is proportional to the number of STPs at a given elevation with the same value of "N". A graph like this tells me quite a lot, such as how much the blow count varies at a certain elevation, and how blow count varies with depth.

http://www.SlideRuleEra.net

 

[r13]

A regression line provides meaningful interpretation to the data points collected, and is ease of comparison. Piggyback on SRE's idea, the statistic chart/graph can then be compared/correlated to other known data sets to validate your work.

 

[ATSE]

SRE said it best, but I will try to add a little.
Judgment is better than statistics when:
- an individual has at least 3 decades of sustained, focused experience
- the experience includes feedback (that is, if your work is never evaluated critically over time via follow up, your judgment has not be validated)

Ron's judgment is very different than a junior engineer with 5 years of experience and the internet.
Judgment can be thought of as a senior engineer accessing his huge database of office and field experience (from his mind) and appropriately applying weight to relevant issues.
For the remaining 90% of engineers, statistics that are proven that be reliable by those old engineers with judgment should be the baseline for decisions.

 

[BigH]

One of my favourites by Ralph Peck:
“There are few things of more importance in ensuring quality on a construction job than to have a set of eyes attached to a calibrated brain observing the construction operations”
Since I am working in Tajikistan at the moment, I've translated this message for the national staff:
"Существуют несколько простых вещей, которые являются важными для улучшения качества строительных работ, и это, прежде всего, использовать зрительный орган (глаза), которые прикреплены к функциональному мозгу, наблюдающему за строительными работами”