Standardized P Chart for Short Run Production

M

mmcbhd

I am using standarized P chart for short run production, do I need to remove the "out-of-control" point after investigation and recalculate Z value?

Or do I keep all data points, but recalcuate P bar value (to get latest Z value) whenever there are failures?
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
You would only "remove" out of control points when trying to initially figure out the baseline. Once you have the baseline locked in, you don't change it unless you have a permanent shift following a trend. I am a little confused to your references to a "z-value" . . .
 
M

mmcbhd

Hi Steve,

I am using "normalized p chart", the formular is to calculate Z value,

Z =(p-pbar)/square foot{[pbar(1-pbar)]}.

the control limits for this normalized p chart is +/-3, central line =0.

I know for Sheward control charts, we need to remove uncontrolled points and re-calculate control limits.

But for the "Normalized p chart", the data points are mainly based on historic pbar value, so I am thinking maybe I don't need to remove uncontrolled points. is this correct?

Thank you!
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Again - it depends. Have you established an initial baseline? If so, do NOT throw out future data. Now, when you have points that are out of control, you do need to identify "what happened" and determine if a permanent shift has occurred such that you need to establish a new baseline.

When you are establishing a baseline, yes, it may be desirable to remove points from the basleine, but you should have both a statistical justification, and an operational justification (that is, you know what caused the anomoly).
 
M

mmcbhd

Looks like for "normalized p chart", there're no need to establish a baseline, as control limits are fixed as +/-3. pretty much we are using p value and P bar value to calculate Z value, then we can tell if the process is out of control or not.
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
Looks like for "normalized p chart", there're no need to establish a baseline, as control limits are fixed as +/-3. pretty much we are using p value and P bar value to calculate Z value, then we can tell if the process is out of control or not.

I strongly disagree. You have to calculate a baseline so you may subtract it from your measured percentage, then you have to calculate the standard deviation in order to shift to the number of standard deviations.

A way to visualize this is you are simply taking a "regular" p-chart, printing it on rubber, and stretching, shifting and shrinking the rubber such that the baseline is still a horizontal line, but at zero, and the jagged control limits become horizontal lines. You must go through the same operations to make a z-value chart as a p-chart.
 

Romvill2002

Involved In Discussions
Hi Steve, could you check attached if my p-chart is correct?

medical device in the field with monthly service/repair data from monthly install base. It seems that p_bar denominator is too large for a total install base volume. But actually the accounts in the field only total to approx 5000 devices. Does not make any sense. please advise....thanks


Month Service Install np p_bar(CL) UCL LCL
Jun/15 8 805 0.010 0.024 0.041 0.008
Jul/15 15 855 0.018 0.024 0.040 0.009
Aug/15 18 1006 0.018 0.024 0.039 0.010
Sep/15 18 1176 0.015 0.024 0.038 0.011
Oct/15 29 1353 0.021 0.024 0.037 0.012
Nov/15 27 1575 0.017 0.024 0.036 0.013
Dec/15 46 1879 0.024 0.024 0.035 0.014
Jan/16 48 2027 0.024 0.024 0.035 0.014
Feb/16 41 2295 0.018 0.024 0.034 0.015
Mar/16 63 2830 0.022 0.024 0.033 0.016
Apr/16 53 2954 0.018 0.024 0.033 0.016
May/16 65 3125 0.021 0.024 0.033 0.016
Jun/16 100 3261 0.031 0.024 0.033 0.016
Jul/16 88 3420 0.026 0.024 0.032 0.017
Aug/16 98 3592 0.027 0.024 0.032 0.017
Sep/16 125 3936 0.032 0.024 0.032 0.017
Oct/16 119 4254 0.028 0.024 0.032 0.017
Nov/16 134 4467 0.030 0.024 0.031 0.018
Dec/16 117 4737 0.025 0.024 0.031 0.018
Total 1212 49547
p̅(CL) 0.024
 

Steve Prevette

Deming Disciple
Leader
Super Moderator
No, as I alluded to in the previous post there is a significant shift in the chart that needs to be accounted for. I would suggest the attached version.
 

Attachments

  • P-chart example SS Prevette.xlsx
    16 KB · Views: 197
Top Bottom