1DBRVSTATDIFF(1) User Contributed Perl Documentation DBRVSTATDIFF(1)
2
6 dbrvstatdiff - evaluate statistical differences between two random
7 variables
8
10 dbrvstatdiff [-f format] [-c ConfRating]
11 [-h HypothesizedDifference] m1c sd1c n1c m2c sd2c n2c
12 OR
14 dbrvstatdiff [-f format] [-c ConfRating] m1c n1c m2c n2c
16
18 Produce statistics on the difference of sets of random variables. If a
19 hypothesized difference is given (with "-h"), to does a Student's
20 t-test.
21 Random variables are specified by:
23 "m1c", "m2c"
25 The column names of means of random variables.
26 "sd1c", "sd2c"
28 The column names of standard deviations of random variables.
29 "n1c", "n2c"
31 Counts of number of samples for each random variable
32 These values can be computed with dbcolstats.
34 Creates up to ten new columns:
36 "diff"
38 The difference of RV 2 - RV 1.
39 "diff_pct"
41 The percentage difference (RV2-RV1)/1
42 "diff_conf_{half,low,high}" and "diff_conf_pct_{half,low,high}"
44 The half half confidence intervals and low and high values for
45 absolute and relative confidence.
46 "t_test"
48 The T-test value for the given hypothesized difference.
49 "t_test_result"
51 Given the confidence rating, does the test pass? Will be either
52 "rejected" or "not-rejected".
53 "t_test_break"
55 The hypothesized value that is break-even point for the T-test.
56 "t_test_break_pct"
58 Break-even point as a percent of m1c.
59 Confidence intervals are not printed if standard deviations are not
61 provided. Confidence intervals assume normal distributions with common
62 variances.
63 T-tests are only computed if a hypothesized difference is provided.
65 Hypothesized differences should be proceeded by <=, >=, =. T-tests
66 assume normal distributions with common variances.
67
69 -c FRACTION or --confidence FRACTION
70 Specify FRACTION for the confidence interval. Defaults to 0.95 for
71 a 95% confidence factor (alpha = 0.05).
72 -f FORMAT or --format FORMAT
74 Specify a printf(3)-style format for output statistics. Defaults
75 to "%.5g".
76 -h DIFF or --hypothesis DIFF
78 Specify the hypothesized difference as "DIFF", where "DIFF" is
79 something like "<=0" or ">=0", etc.
80 This module also supports the standard fsdb options:
82 -d Enable debugging output.
84 -i or --input InputSource
86 Read from InputSource, typically a file name, or "-" for standard
87 input, or (if in Perl) a IO::Handle, Fsdb::IO or Fsdb::BoundedQueue
88 objects.
89 -o or --output OutputDestination
91 Write to OutputDestination, typically a file name, or "-" for
92 standard output, or (if in Perl) a IO::Handle, Fsdb::IO or
93 Fsdb::BoundedQueue objects.
94 --autorun or --noautorun
96 By default, programs process automatically, but Fsdb::Filter
97 objects in Perl do not run until you invoke the run() method. The
98 "--(no)autorun" option controls that behavior within Perl.
99 --help
101 Show help.
102 --man
104 Show full manual.
105
107 Input:
108 #fsdb title mean2 stddev2 n2 mean1 stddev1 n1
109 example6.12 0.17 0.0020 5 0.22 0.0010 4
110 Command:
112 cat data.fsdb | dbrvstatdiff mean2 stddev2 n2 mean1 stddev1 n1
113 Output:
115 #fsdb title mean2 stddev2 n2 mean1 stddev1 n1 diff diff_pct diff_conf_half diff_conf_low diff_conf_high diff_conf_pct_half diff_conf_pct_low diff_conf_pct_high
116 example6.12 0.17 0.0020 5 0.22 0.0010 4 0.05 29.412 0.0026138 0.047386 0.052614 1.5375 27.874 30.949
117 # | dbrvstatdiff mean2 stddev2 n2 mean1 stddev1 n1
118 Input 2:
120 (example 7.10 from Scheaffer and McClave):
121 #fsdb title x2 sd2 n2 x1 sd1 n1
123 example7.10 9 35.22 24.44 9 31.56 20.03
124 Command 2:
126 dbrvstatdiff -h '<=0' x2 sd2 n2 x1 sd1 n1
127 Output 2:
129 #fsdb title n1 x1 sd1 n2 x2 sd2 diff diff_pct diff_conf_half diff_conf_low diff_conf_high diff_conf_pct_half diff_conf_pct_low diff_conf_pct_high t_test t_test_result
130 example7.10 9 35.22 24.44 9 31.56 20.03 3.66 0.11597 4.7125 -1.0525 8.3725 0.14932 -0.033348 0.26529 1.6465 not-rejected
131 # | /global/us/edu/ucla/cs/ficus/users/johnh/BIN/DB/dbrvstatdiff -h <=0 x2 sd2 n2 x1 sd1 n1
132 Case 3:
134 A common use case is to have one file with a set of trials from two
135 experiments, and to use dbrvstatdiff to see if they are different.
136 Input 3:
138 #fsdb case trial value
140 a 1 1
141 a 2 1.1
142 a 3 0.9
143 a 4 1
144 a 5 1.1
145 b 1 2
146 b 2 2.1
147 b 3 1.9
148 b 4 2
149 b 5 1.9
150 Command 3:
152 cat two_trial.fsdb |
153 dbmultistats -k case value |
154 dbcolcopylast mean stddev n |
155 dbrow '_case eq "b"' |
156 dbrvstatdiff -h '=0' mean stddev n copylast_mean copylast_stddev copylast_n |
157 dblistize
158 Output 3:
160 #fsdb -R C case mean stddev pct_rsd conf_range conf_low conf_high conf_pct sum sum_squared min max n copylast_mean copylast_stddev copylast_n diff diff_pct diff_conf_half diff_conf_low diff_conf_high diff_conf_pct_half diff_conf_pct_low diff_conf_pct_high t_test t_test_result t_test_break t_test_break_pct
162 case: b
163 mean: 1.98
164 stddev: 0.083666
165 pct_rsd: 4.2256
166 conf_range: 0.10387
167 conf_low: 1.8761
168 conf_high: 2.0839
169 conf_pct: 0.95
170 sum: 9.9
171 sum_squared: 19.63
172 min: 1.9
173 max: 2.1
174 n: 5
175 copylast_mean: 1.02
176 copylast_stddev: 0.083666
177 copylast_n: 5
178 diff: -0.96
179 diff_pct: -48.485
180 diff_conf_half: 0.12202
181 diff_conf_low: -1.082
182 diff_conf_high: -0.83798
183 diff_conf_pct_half: 6.1627
184 diff_conf_pct_low: -54.648
185 diff_conf_pct_high: -42.322
186 t_test: -18.142
187 t_test_result: rejected
188 t_test_break: -1.082
189 t_test_break_pct: -54.648
190 # | dbmultistats -k case value
192 # | dbcolcopylast mean stddev n
193 # | dbrow _case eq "b"
194 # | dbrvstatdiff -h =0 mean stddev n copylast_mean copylast_stddev copylast_n
195 # | dbfilealter -R C
196 (So one cannot say that they are statistically equal.)
198
200 Fsdb. dbcolstats. dbcolcopylast.
201
203 Copyright (C) 1991-2018 by John Heidemann <johnh@isi.edu>
204 This program is distributed under terms of the GNU general public
206 license, version 2. See the file COPYING with the distribution for
207 details.
208perl v5.32.1 2021-01-27 DBRVSTATDIFF(1)