Anda di halaman 1dari 10

Basic

Info:
606 Unique respondants
Hypothesis:
1) The long deadline conditions should have higher hits than the no deadline condition (No
difference in false alarms expected)
2) The short deadline condition should have higher hits than the long deadline condition (No
difference in false alarms expected)
No deadline:
Reduces false alarms and also hits
1000ms Deadline:

500ms Deadline:
















Descriptive statistics:

[The confidence interval is not a descriptive statistic]


[The confidence interval 95% is that if you were to sample many times 95% of those
times the mean would be in the ___???]
[ignore the skew in the distribution]
[Ignore Kurtosis in the distribution]
[Condition 1: No
Condition 2: 1000ms
Condition 3: 500ms

Con
Case Processing Summary
Cases
Valid
Con
Hits

FAs

Missing
Percent

Total

Percent

Percent

181

100.0%

0.0%

181

100.0%

183

100.0%

0.0%

183

100.0%

193

100.0%

0.0%

193

100.0%

181

100.0%

0.0%

181

100.0%

183

100.0%

0.0%

183

100.0%

193

100.0%

0.0%

193

100.0%

Descriptives
Con
Hits

Statistic
Mean
95% Confidence Interval for

Lower Bound

.7951

Mean

Upper Bound

.8221

5% Trimmed Mean

.8128

Median

.8200

Variance
Std. Deviation

.09224
.56

Maximum

.98

Range

.42

Interquartile Range

.12

Kurtosis
Mean

.00686

.009

Minimum

Skewness

.8086

Std. Error

-.611

.181

.054

.359

.8055

.00714

95% Confidence Interval for

Lower Bound

.7914

Mean

Upper Bound

.8196

5% Trimmed Mean

.8093

Median

.8200

Variance

.009

Std. Deviation

.09658

Minimum

.50

Maximum

.98

Range

.48

Interquartile Range

.14

Skewness

-.596

.180

Kurtosis

-.115

.357

Mean

.7848

.00730

95% Confidence Interval for

Lower Bound

.7704

Mean

Upper Bound

.7992

5% Trimmed Mean

.7889

Median

.8000

Variance

.010

Std. Deviation

.10148

Minimum

.48

Maximum

.98

Range

.50

Interquartile Range

.14

Skewness
Kurtosis
FAs

Mean

-.663

.175

.268

.348

.1862

.00752

[False

95% Confidence Interval for

Lower Bound

.1713

alarm

Mean

Upper Bound

.2010

rates]

5% Trimmed Mean

.1808

Median

.1800

Variance
Std. Deviation

.010
.10118

Minimum

.02

Maximum

.52

Range

.50

Interquartile Range

.13

Skewness

.848

.181

Kurtosis

.436

.359

Mean

.1628

95% Confidence Interval for

Lower Bound

.1497

Mean

Upper Bound

.1759

5% Trimmed Mean

.1592

Median

.1600

Variance

.008

Std. Deviation

.08979

Minimum

.00

Maximum

.44

Range

.44

Interquartile Range

.14

Skewness

.00664

.446

.180

Kurtosis

-.212

.357

Mean

.2172

.00708

95% Confidence Interval for

Lower Bound

.2032

Mean

Upper Bound

.2312

5% Trimmed Mean

.2142

Median

.2000

Variance
Std. Deviation

.010
.09842

Minimum

.02

Maximum

.58

Range

.56

Interquartile Range

.13

Skewness

.565

.175

Kurtosis

.321

.348

Graph


One Way ANOVA

Tells us: ANOVA tells us that there is a significant effect of our line-up manipulation among
our groups

Oneway
ANOVA
Sum of Squares
False Alarms

Hits

Between Groups

df

Mean Square

.280

.140

Within Groups

5.170

554

.009

Total

5.450

556

.063

.032

Within Groups

5.206

554

.009

Total

5.270

556

Between Groups

Used to calculate the F value, significance (P value).


F(2, 554) = 3.38, p = .035
F(2, 554) = 15.01, p < .001

p = .035

Effect Size: (How big the effect is) 2
Effect size for hits: .063/5.27 = .01
Effect size for false alarms: .28/5.45 = .05
What do those numbers mean?

Cohen (1988) [reference on LMS] suggests the following rule of thumb:

0.0099 constitutes a small effect

0.0588 a medium effect

0.1379 a large effect

When is the effect: Bonferroni corrections

Sig.

15.009

.000

3.375

.035

Post Hoc Tests


Multiple Comparisons
Bonferroni
Mean Difference

95% Confidence Interval

(I-J)
[report the
Dependent Variable

(I) Condition

(J) Condition

False Alarms

No Deadline

1000ms

1000ms

500ms

Hits

No Deadline

1000ms

Std. Error

Sig.

Lower Bound

Upper Bound

.02335

.01013

.065

-.0010

.0477

500ms

-.03101

.01000

.006

-.0550

-.0070

No Deadline

-.02335

.01013

.065

-.0477

.0010

500ms

-.05436

.00997

.000

-.0783

-.0304

No Deadline

.03101

.01000

.006

.0070

.0550

1000ms

.05436

.00997

.000

.0304

.0783

1000ms

.00315

.01016

1.000

-.0212

.0276

500ms

.02385

.01003

.053

-.0002

.0479

-.00315

.01016

1.000

-.0276

.0212

.02070

.01000

.117

-.0033

.0447

No Deadline

-.02385

.01003

.053

-.0479

.0002

1000ms

-.02070

.01000

.117

-.0447

.0033

No Deadline
500ms

500ms

absolute value]

*. The mean difference is significant at the 0.05 level.

How to report a post-hoc test


A post-hoc analysis using the Bonferroni correction for multiple comparisons showed no
significant difference in the hit rate between any of the conditions, but a significant difference
between the false alarm rates for the No Deadline condition and the Short Deadline condition
(Mean Difference = .03, p = .006), and between the Long Deadline condition and the Short
Deadline conditions (Mean Difference = .05, p < .001).


Why is the ANOVA significant for Hit Rate but the Bonferroni test is not?

The Bonferroni test is a more conservative test because it adjusts the estimated p-value for
the number of comparisons that we are making

Our conclusion would have to be that the significant ANOVA is likely a false positive

Report both, but point out the above, saying that the B test is the betterer

Anda mungkin juga menyukai