Project Report An investigation on data entry effectiveness & efficiency
Dec 14, 2015
Project ReportAn investigation on data entry effectiveness & efficiency
AgendaIntroduction
Experiment Design
Experiment Platform
Data Analysis
Conclusion
Introduction Developing fast and efficient means
for text entry is one of the most pressing research topics in today!
Although the QWERTY layout is entrenched for physical keyboards, soft keyboards are easy to implement and customize.
Objective To develop a statistically reliable
analysis of possible factors that may have influence on the input efficiency and error rateof the soft keyboard and give design guideline on this issue.
Response variables
Average typing speed
Error rate
Cause–and-effect Diagram
Control variablesKeyboard layout
Key size
Key shape
Mouse C/R ratio
Keyboard layout
Other Control FactorsKey size Large: the size of keys on
standard physical keyboard
Small: that of general electronic dictionaries
Mouse C/R ratio
High: the highest in the tested computer.
Low: adjusted one which is the lower bound that one can bear
Key shape There are three
kinds of key shape generally, but we choose two of them for simplicity: circle and square, the rest is hexagon
Learning Curve for Novice
Experiment Design(1/3)Mixed Level Full Factorial Design
Repleicated on two testeesEach testee enters the same
paragraph of text with different combinations of control factors
Experiment Design(2/3)
Experiment Design(3/3)Block On Noise Variables
Data within each block is homogenous (exclude the potential influence of different users)
Enough DOF to Estimate Interactions
Eliminate Learning Effect Two replicates with different (randomly generated)
execution sequence
Platform(1/4)Large-SmallHigh-LowFITALY-OPTI II-
METROPOLIS-QWERTY Square-Circle.
There are 16 different types of soft keyboard in all.
Platform(2/4)
FITALY OPTI II
METROPOLIS QWERTY
Platform(3/4)
Platform(4/4)Saved in CSV files
Key name, x-coordinate, y-coordinate, the time when it is entered
Errors: number of “Backspace”
Discussion on Error RateErrors may not be noticed immediately.
Sequence of the words or letters may be changed, due to other reasons unrelated to the keyboard.
Click “Backspace” when ever testee notices error and count the number of “Backspace” clicked as error
AnalysisTools
DOE++Minitab
Two Kinds of ResponsesTimeError rate
PrincipleTreat different responses separatelyFind an integrative optimal solution
Response TimeData Analysis
Step 1: ANOVAThe main effect is
significant
Step 2: RegressionIgnore the 3 or
higher order effectsKeep the significant
items
Step 3:Probability PlotReliaSoft DOE++ - www.ReliaSoft.com
Normal Probability Plot of Effect
Alpha = .1
Standardized Effect (T Value)
Pro
ba
bili
ty
-6.000 6.000-3.600 -1.200 1.200 3.6001.000
5.000
10.000
50.000
99.000
B[1]
B[2]
B[2]CD
A:Size
B[3]
Effect Probability
TimeY' = Y
Non-Significant EffectsSignificant EffectsDistribution Line
宁毅清华大学2008-6-810:26:49
Step 4: Eliminate Insignificant factors
Keep A, B1, B2, B3
Step 5: Residual Checking
Step 6: Interaction Matrix
Conclusion
Time will be shorter whenfactor size= 2(larger size)factor layout=1(qwerty), and then 4(FITALY),
2(metro), 3(OPTI)The factor shape and C/R ratio will not influence
the result if the response is time
Error RateData Analysis
Step 1: ANOVAThe main effect is
significant
Step 2: RegressionIgnore the 3 or higher
order effectsKeep the significant
items
Step 3: Probability Plot
Step 4: Eliminate Insignificant Factors
Keep B1, AB1, AB2, C
Step 5: Residual Checking
Step 6: Interaction Matrix
Conclusion
It is more reasonable to choose factor size= 2(larger size)
Hence factor layout = 4(FITALY) will be chosen when the size=2
The factor C/R ratio=1(low) will be betterFactor shape has no effects
Final RecommendationLarger size
FITALY layout
Low C/R ratio
Both shapes
Potential WeaknessLimited Sample Size
Both Young Female University Students
Familiar with QWERTY
Without Upper Case
Limited to English Text
Q&A
Thank you