| |
|
|
Regression
Analysis |
|
Project |
Our client, a
regional manufacturer, approached SQPS with a request to
analyze historical data. From the analysis, our client was
interested in identifying relationships among variables and
to be able to predict some response variables given the
values of independent variables. There were 21 independent
variables and a response (Throughput). |
|
Approach |
Our approach
can be summarized as follows:
- A team
consisting of the client's continuous improvement
associates was assembled
-
Historical data was reviewed to ensure accuracy,
validity and completeness
- Some
data transformation was conducted
-
Correlation analysis was performed to identify
relationships among independent variables
-
Stepwise regression was run to develop a prediction
formula
-
Prediction formula was verified using other raw data
|
|
Results |
The results
showed that
-
Relationships among independent variables were
identified.
- highly
correlated variables were examined as we only needed one
of each pair for the prediction model
- A
prediction model for "Throughput" was estimated with
R-Sq of 88%. This means that 88% of the variability in
the "Throughput" data was explained by the effect of the
model. This was a very reliable prediction model.
|
|
|
| |
|
|
|
|
|
|
