Please enable JavaScript.
Coggle requires JavaScript to display documents.
IFN654: Week 6 - Analyse Phase I - Coggle Diagram
IFN654: Week 6 - Analyse Phase I
Wrap up Measure Phase
Probable input variable X's causes Y
collect required process data
measure process capability
What is Analyze Phase purpose? :star:
Analyzed collected data, using graphical tools, to prove/disprove the impact of Xs on Y
I . "X" Sifting - Screening
Define Phase -> Sipoc -> Possbile X's
Measure Phase -> Process Map/ C E (X-Y) Matrix, FMEA -> Probable X's
Multi-Vari Analysis :tada:
display patterns of variation
identify families of variation
three families of var exist within a subgroup/ between subgroups/ or over time
Within a unit - Positional - Across a single unit contains many individual parts
Understand whats happening in the sample mean
Between a unit - Cyclical - among consecutive peices
Looking at the variation between different collected data samples
Over time - Temporal - Shift to shift/ Day to day/ week to week
Looking at time series analysis
Explain graphical relationship between X and Y
Multi-Vari Method :star:
Creating Sampling plan (covering Within, Between, and Temporal) - represents 80% of variation in the process :star2:
Gather (additional) Data
Graph Data
Check to see if Variation is Exposed
Interpret Results
Classes and Causes of Distributions :tada:
Normal Distribution
most important distribution
area under curve = 1
mounded/symmetric
Pure random variation
Mean = 0 and ST = 1
Mean, Median, Mode = same data point
Special Causes/ Common Causes :forbidden:
Special Causes: Variation caused by Known factors -> non-random distribution of outpt (Assignable Cause)
Common Causes: Variation caused by unknown factors -> random distribution of output around the average of data (left over after remove Special Causes)
Non-normal Distribution :+1:
Skewness
Distribution longest tail points in direction of skew
Potential causes
Natural Limits
Artificial Limits (Sorting)
Mixtures
Non-Linear Relationships
Interaction
Non-random Patterns across time
Moment coefficient of Skewness close to 0 (for normal distribution)
Kurtosis
Refer to shape:
Peaked with long tail (Leptokurtic)
Flat with short tail (Platykurtic)
Platykurtic
Multiple Means shifting over time :star:
Kurtosis value = negative
p-value (kurtosis) = 0.000 -> significantly confirms the flatness of data
Lepokurtic
Distributions overlaying each other :star:
Kurtosis value = positive
p-value (kurotisis) < 0.05 significantly confirms the lepokurtic distribution
Multiple Modes
Mixture of distributions
Bimodal Distributions
Two distributions within a data
2 different machines/ operators/ administrators
Extreme Bi-modal (outliers)
Own-variation easily assigned
Bi-modal (multiple outliers)
Multiple inputs into the process
Q1/Q2/Q3 near minimum
High Skewness/ kurtosis/ non normal from A-D test -> indicates two different modes :star:
Grannular
UNDERSTAND THE DATA, WHAT IS THE STORY, AND WHAT THE DATA IS TELLING US :star:
Inferential Statistics
5 steps approach
What do you want to know?
What tool will give you that information?
What kind of data does that tool require?
How will you collect the data?
How confident are you with your data summaries?
Types of error
Error in sampling
Bias in sampling
Error in measurement
Lack of measurement validity
Populations/ Sample/ Observation
Population: Every data point
Sample: A portion/ subset of population
Observation: Individual measurement
Observations
Sample size increase
Center remains the same
2, Variation decreases
Shape of distribution changes -> tend to be normal
What is a good sample size?
sample size of 30 is generate summary statistics such as Mean and Standard Deviation
To check the probability of your results applying to wider population :star: