Experiments - Samuel marden

advertisement
Experiments
I will try and post the slides used in
class each week on my website
http://samuelmarden.weebly.com/
(a) What is the experimental setup?
(a) What is the experimental setup?
1. Approaches people at a sportscard convention and
asks if they’d be willing to complete a survey.
2. Randomly gives accepters either ‘Card A’ or ‘Card B’.
–
–
Why is it important that it is random?
Why is it important that they give the card after people
have accepted.?
3. After they complete the survey give them the
opportunity to trade
Why should we expect exactly half of people to trade?
(b) What Percentage of people did trade?
use "data-exercise-10-sportscards", clear
. sum
Variable |
Obs
Mean
Std. Dev.
Min
Max
-------------+-------------------------------------------------------trade |
148
.3378378
.474579
0
1
trades_p_m |
148
10.24324
10.10127
0
70
years_trade |
148
8.655405
8.316006
0
60
dealer |
148
.5
.5016978
0
1
goodb |
148
.527027
.5009643
0
1
-------------+-------------------------------------------------------income |
148
4.148649
1.987597
1
8
education |
148
3.628378
1.467472
1
6
male |
148
.8986486
.3028182
0
1
age |
148
34.68919
13.01849
10
76
Is this percentage statistically different from 50%?
1.
Specify the null and alternative hypothesis.
2.
Specify the confidence interval and rejection region
3.
Can we perform the test with the information here?
(c) Is there any evidence that one of the two
‘sportscards’ was more popular?
What test shall we run?
(c) Is there any evidence that one of the two
‘sportscards’ was more popular?
ttest trade, by(goodb)
Two-sample t test with equal variances
-----------------------------------------------------------------------------Group |
Obs
Mean
Std. Err.
Std. Dev.
[95% Conf. Interval]
---------+-------------------------------------------------------------------0 |
70
.3285714
.0565445
.4730851
.2157683
.4413746
1 |
78
.3461538
.0542159
.4788222
.2381961
.4541116
---------+-------------------------------------------------------------------combined |
148
.3378378
.0390102
.474579
.2607447
.414931
---------+-------------------------------------------------------------------diff |
-.0175824
.0783882
-.1725045
.1373397
-----------------------------------------------------------------------------diff = mean(0) - mean(1)
t = -0.2243
Ho: diff = 0
degrees of freedom =
146
Ha: diff < 0
Pr(T < t) = 0.4114
Ha: diff != 0
Pr(|T| > |t|) = 0.8228
Ha: diff > 0
Pr(T > t) = 0.5886
Although it doesn’t matter for the validity of the experiment, it is
helpful that the two items seem equally popular. Why?
(c) Is there an endowment effect amongst
dealers?
What test shall we run?
(c) Is there an endowment effect amongst
dealers?
ttest trade == .50 if dealer == 1
One-sample t test
-----------------------------------------------------------------------------Variable |
Obs
Mean
Std. Err.
Std. Dev.
[95% Conf. Interval]
---------+-------------------------------------------------------------------trade |
74
.4459459
.0581776
.5004626
.3299982
.5618937
-----------------------------------------------------------------------------mean = mean(trade)
t = -0.9291
Ho: mean = .50
degrees of freedom =
73
Ha: mean < .50
Pr(T < t) = 0.1779
Ha: mean != .50
Pr(|T| > |t|) = 0.3559
Ha: mean > .50
Pr(T > t) = 0.8221
(e) Are the measures of experience in the data
set correlated?
corr dealer trades_p_m years_trade
(obs=148)
|
dealer trades~m years_~e
-------------+--------------------------dealer |
1.0000
trades_p_m |
0.4551
1.0000
years_trade |
0.2063
0.0400
1.0000
(e) Are the measures of experience in the data
set correlated?
pwcorr dealer trades_p_m years_trade, sig
|
dealer trades~m years_~e
-------------+--------------------------dealer |
1.0000
|
|
trades_p_m |
0.4551
1.0000
|
0.0000
|
years_trade |
0.2063
0.0400
1.0000
|
0.0119
0.6290
|
(e) Do people with above average experience
trade more often?
What test shall we run?
Which average should we use?
kdensity trades_p_m
Mean=10.2, median=8
sum trades_p_m, d
ttest trade==0.5 if trades_p_m>r(p50)
One-sample t test
-----------------------------------------------------------------------------Variable |
Obs
Mean
Std. Err.
Std. Dev.
[95% Conf. Interval]
---------+-------------------------------------------------------------------trade |
69
.4927536
.0606275
.5036102
.3717733
.613734
-----------------------------------------------------------------------------mean = mean(trade)
t = -0.1195
Ho: mean = 0.5
degrees of freedom =
68
Ha: mean < 0.5
Pr(T < t) = 0.4526
Ha: mean != 0.5
Pr(|T| > |t|) = 0.9052
Ha: mean > 0.5
Pr(T > t) = 0.5474
sum trades_p_m, d
ttest trade==0.5 if trades_p_m>r(mean)
One-sample t test
-----------------------------------------------------------------------------Variable |
Obs
Mean
Std. Err.
Std. Dev.
[95% Conf. Interval]
---------+-------------------------------------------------------------------trade |
49
.5102041
.0721538
.5050763
.3651292
.6552789
-----------------------------------------------------------------------------mean = mean(trade)
t =
0.1414
Ho: mean = 0.5
degrees of freedom =
48
Ha: mean < 0.5
Pr(T < t) = 0.5559
Ha: mean != 0.5
Pr(|T| > |t|) = 0.8881
Ha: mean > 0.5
Pr(T > t) = 0.4441
sum years_trade, d
ttest trade==0.5 if years_trade>r(p50)
One-sample t test
-----------------------------------------------------------------------------Variable |
Obs
Mean
Std. Err.
Std. Dev.
[95% Conf. Interval]
---------+-------------------------------------------------------------------trade |
71
.3802817
.0580231
.4889112
.2645583
.4960051
-----------------------------------------------------------------------------mean = mean(trade)
t = -2.0633
Ho: mean = 0.5
degrees of freedom =
70
Ha: mean < 0.5
Pr(T < t) = 0.0214
Ha: mean != 0.5
Pr(|T| > |t|) = 0.0428
Ha: mean > 0.5
Pr(T > t) = 0.9786
What do we conclude about the effect of
experience?
f) Regress the dummy
variable trade on one of
the measures of trading
experience and these
additional control
variables using both OLS
and probit models. What
do you conclude from the
results?
reg trade dealer age male income
education, robust
probit trade dealer age male income
education
(1)
(2)
VARIABLES
trade
trade
OLS
Probit
dealer
0.228***
0.646***
(0.0789)
(0.225)
age
-0.00239
-0.00778
(0.00311)
(0.0107)
male
0.107
0.370
(0.104)
(0.408)
income
-0.0157
-0.0431
(0.0224)
(0.0702)
education
0.0362
0.102
(0.0288)
(0.0844)
Constant
0.145
-1.025**
(0.170)
(0.503)
Observations
148
148
R-squared
0.072
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Probit with marginal effects calculated. Where are they calculated?
VARIABLES
dealer
age
male
income
education
(1)
trade
0.228***
(0.0789)
-0.00239
(0.00311)
0.107
(0.104)
-0.0157
(0.0224)
0.0362
(0.0288)
(2)
trade
Probit
0.230***
(0.0776)
-0.00281
(0.00385)
0.123
(0.123)
-0.0156
(0.0253)
0.0367
(0.0304)
trades_p_m
(3)
trade
(4)
trade
Probit
(5)
trade
(6)
trade
Probit
-0.00136
(0.00325)
0.129
(0.0936)
-0.00869
(0.0226)
0.0178
(0.0283)
0.0135***
(0.00340)
-0.00155
(0.00382)
0.152
(0.118)
-0.00817
(0.0252)
0.0169
(0.0298)
0.0147***
(0.00441)
-0.00211
(0.00385)
0.155
(0.112)
-0.00910
(0.0235)
0.0201
(0.0292)
-0.00230
(0.00398)
0.159
(0.113)
-0.00925
(0.0247)
0.0191
(0.0296)
-0.00101
(0.00445)
0.245
(0.181)
-0.00112
(0.00539)
148
0.017
148
years_trade
Constant
0.145
(0.170)
Observations
148
148
R-squared
0.072
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
0.102
(0.154)
148
0.099
148
(g) Are there any apparent problems with the
internal validity of the results that we have
obtained?
What conclusions would you draw from this
experiment for studies that are trying to
evaluate how much people are willing to pay for
a particular intervention, such as, for example, a
nature conservation project?
Download