It certainly is a senior management issue. Where I work, the CEO:
– sent out an e-mail that Six Sigma and DFSS was the way we operate (he always gives the same message in person)
– nobody will be promoted before they take SS training (this gets senior people signed up very quickly)
– anybody blocking the process will be kicking their…[Read more]
NS replied to the topic Two Way Anova Problem ! Interaction Output missing in the forum General 13 years, 3 months ago
It does not look like you have enough degrees of freedom. You have 2 df for tenure (3-1) and 6 df for qualification (7-1). The intereaction would take 6 x 2 = 12 df, giving a total of 20 df. You only have 21 data points (or 20 df).
This is the way that I look at it. In most of our plants, 5-10% of our activites are Value-Added. Yet, we spend a great deal of time trying to improve/optimize the 5% VA steps.
With a Process Analysis or Transactional approach you get the chance to directly impact 90-95% of the process. These projects usually yield a substantial amount of…[Read more]
These are NVA activites. We do not have to stock them or give the item a grand tour of the store.
You also answered your own question. The task of bringing the item to the store is also NVA. The item could have been shipped directly to my house.
For a continuous X and a continuous Y, you should do a BLR (Binary Logistic Regression).
What you are referring to is a Reverse ANOVA. You switch the “X” and the “Y”. Since you are establishing correlation (as opposed to causality), the reverse ANOVA will tell you if there is a relationship.
I agree with Stan, less so with Ray.
We train our Senior Managers at a high level (what is a good project, how to choose projects, Project Charters, Six Sigma Philosophy, CTS Flowdowns, quick overview of the 12 steps). We concentrate on the Recognize and Define phases. This takes 3-4 days.
No need to teach Senior Management about p-values or h…[Read more]
The “observed” performance uses the actual data points (over the USL and below the LSL).
The “expected” uses the “normal curve” that Minitab generates (if you selected a “normal” distribution). Minitab calculates the “Z” score for the defects that fall to the right of the USL, and a “Z” score for the portion of the curve to the left of the…[Read more]
R2 (adj) corrects for the number of terms in the equation and the number of data points. It is calculated using:
R2 (adj) = 1 – (n-1)/(n-p) * (1-R^2)
where n – number of observations (data points) and p = total number of terms (including the constant)
Practically, it tells you how good the model is. For example, is a quadratic model better t…[Read more]
There were a few strings about this. Start with:
No Black or White answer.
A C&E will focus on what “caused” the failure “after the fact”. That is why it is commonly associated with a RCFA analysis.
An FMEA is more “pro-active”, it asks the question: how can this fail (before it actually fails)? More common with a RCM approach. It can be more time consuming because it makes you think abo…[Read more]
Say you had a continuous in-line flowmeter for natural gas.
Set the reading at 1000. Measure a “burst (20?)” of short-term readings. The time-frame would be very short, and depends on the instrumentation in your plant (how fast you measure and record). Do the same for 1500, 2000, 2500, etc. You go to Minitab, parts=levels that you chos…[Read more]
You are mixing up “accuracy” with “precision”. Calibration will solve the first issue, and should be done prior to the Gage.
Of course you can do a Gage R&R on continuous measurements. You have to make the assumption that for “short” spurts or periods of time that the reading is stable (ie. the unit is at the same setting, or gives the same r…[Read more]
This is a rule of thumb. As Darth said, no strict rules that I know of.
Treat discrete as continuous if:
– data is numerical (obvious!)
– at least 20 “buckets”
– distance between each “bucket” is the same
– data is normally distributed
Andrea – 1037 would be the required responses that you need to receive.
This site shows the calculation in more detail.
There are lots of sites that have “statistic” quotes:
– there is no influence of parts (all the parts are the same)
– all the operators behave the same
– no effect of operator/part ineraction
The null hypo is that there is no effect, for each of the 3 cases. If p<0.05 (usual 95%), then you have found an effect. In general, we want an effect due to the parts (if you have chosen a wi…[Read more]
Here is an example that will hopefully illustrate the point:
(1) You have a group of people. You measure the level of oxygen in their bloodstream. You then give each person a pill. You re-measure the level of the oxygen in their bloodstream. Here it would be valid to use a “paired t-test” as you can match “before and after” for person A, fo…[Read more]
There have been many posts on this subject in the past. Type in safety in the “top right hand corner” and you will get lots of information.
- Load More