Erik L

Ram,
 
The rules that were developed to employ control charts as a means of process control were designed to strike an economic balance.  The decision rule to say that something had changed/or an issue was developing that should be addressed was set at a level that would minimize jumping after noise/phantoms in the process (alpha risk) yet s…[Read more]

Vimal,
 
There are a couple of ways that you could proceed here.  I’ll go over the passive approaches first.  The easiest, although doesn’t seem realistic for your scenario, would be if there is a simple linear regression that has been established between your x and Y.  If that was true, you could specify a prediction interval around the mean…[Read more]

George,
dof in contingency tables is determined by the minimum number of cells that must be filled to be able to calculate the remainder, given the summations for the rows and columns.
Regards,
Erik

You can always take the approach that Wheeler has suggested.  Observational/experimental studies with small values triggering statistical signifiicance (.05 and lower) are practically stating a desire to maintain the status quo and adjustments/signals of deviation have to be very large to change the null that everything is good in the ‘hood (ala…[Read more]

Gabriela,
I have another suggestion of how you can take the data that you currently have and derive the assessment of the variation attributed to measure.  With this approach you’ll be able to generate a graphical depiction of the various variance components and assess this from a practical perspective.  To achieve this you’ll need to str…[Read more]

DNB,
 
Based off of the scenario that you’ve described I would suggest the following course of action.  You should be able to assess the validity of your sample size by making some statements around when you are going to state statistical significance and the desired power that you would like to detect the practical difference in scores be…[Read more]

Andrew,
There’s a case study application within Understanding Industrial Designed Experiments, by Schmidt and Launsby relative to simulation within SDI.  Might provide some of the insight that you are looking for.
Regards,
Erik

pete,
If you’d like a very good general reference text, I highly encourage that you look into Applied Linear Statistical Models, by Neter, Kutner, et al.  This is a very complete coverage with many examples of data sets along the lines that you are interested in pursuing. 
Another item to throw into the mix.  Is going to be assessing the st…[Read more]

satu,
I’ll try the ‘teach to fish’ approach here.  Any hypothesis testing scenario can be positively influenced by following this roadmap. 
First, what is the x and what is the Y?  What data type are you dealing with?  Is the measurement system believable?  What is considered a practical difference of interest?  Ok, prereqs out of the way. …[Read more]

Sniper,
 
I think it’s ludicrous advice to say that the gage doesn’t have impact on the validity of the control chart.  Measurement error can magnify the likelihood of committing either alpha or beta errors-effectively neutralizing the benefit of employing SPC to status the VOP.  You can remind your MBB that the physical value ‘observed’ is a s…[Read more]

Jeff,
Another thought might be to regress age of the employee against time of service.  If there wasn’t anything there, the grand average of the response would be as good a predictor as the independent variable hypothesized….If there is a relationship, the Pareto/Chi-square analysis suggested could point to likely culprits behind the…[Read more]

fresher,
 
The key to the appropriate application of any tool is to first understand the question that is to be answered.  What is the current hypothesis that is being assessed with the control chart?  I’m hoping, based off of your initial message, that there have already been some discussions around what within your process constitutes a ‘ra…[Read more]

Converting data from a normally distributed data set to a standard normal distribution allows for one table of probabilities to cover an infinite number of potential real-world scenarios.  For simplicity, one could imagine for every combination of N(Xbar, var), there would be an associated table that would provide cumulative probabilities, as…[Read more]

Michael,
In my experiences, the three primary reasons that studies/experiments fail in manufacturing are due to:  not conducting MSAs-Gage R&Rs; lack of attention to power and sample size (including what the business understands is ‘practically significant signal’ ; and not taking the time to fundamentally understand the true issue and what is…[Read more]

Claude,
It would probably be insightful to see what values are being created within your capability calculations.  But, if the assumption is that you are tracking “defectives” and not “defects”, we don’t need the opportuntiy count to derive the numbers for PPM, Cpk, or Sigma level.  Hope that helps shed some light.
Regards,
Erik

Joe,
It will always be easier to reject the null, and assume the alternative state with a directional alternative vs. a not equal alternative.  The best way to think about it would be to take the significance level you are interested in and then partition it against a reference null distribution.  Assuming 5% significance, and a one sided a…[Read more]

DM,
 
Sorry if my response caused further confusion.  To make the calculations work for a minimum sample size requires that certain pieces of information be provided.  Since I had no baseline to work from, I made an assumption that an estimate of your processes baseline proportion defective was 50%.  This can be changed to whatever you feel is…[Read more]

DM,
 
There are a couple of items that will be needed to get you to an estimate of the minimal sample size required.  I don’t know if you have Minitab and/or JMP, so we can use Excel to get there… 
 
One item that is needed is an estimate of the baseline proportion defective.  Since there wasn’t one specified I’ll use 50% (We’ll call this…[Read more]

DM,
 
There are a couple of items that will be needed to get you to an estimate of the minimal sample size required.  I don’t know if you have Minitab and/or JMP, so we can use Excel to get there… 
 
One item that is needed is an estimate of the baseline proportion defective.  Since there wasn’t one specified I’ll use 50% (We’ll call this…[Read more]