Dear Stata community,
is there a post-hoc test after a significant Friedman test that also accounts for repeated measurements? By the Friedman test, I am referring to the user written command -friedman-.
In the literature I have read that the Dunn-Bonferroni test (in Stata implemented by user written command -dunntest-) is the appropriate post-hoc test for the Friedman test. However, I am unsure whether this test accounts for or lets the user account for repeated measurements the same way the Friedman test does. When studying the documentation on -dunntest-, I could not find any indication that this is the case.
A brief elaboration on my research: I have 30 subjects that are measured four times, where each measurement point is several months apart. I want to test whether there is a significant change in the measurements on subject-level across time.
My data does not follow a normal distribution and I want to account for the fact that the same subjects are measured repeatedly (i.e. that I have dependent groups). Hence, I conducted the Friedman test to test for differences between groups (i.e., the measurement times m1 to m4). The Friedman test was significant.
I would now like to conduct an appropriate post-hoc test which also accounts for the fact that I have repeated measurements.
Thank you very much.
Best,
Tobias
is there a post-hoc test after a significant Friedman test that also accounts for repeated measurements? By the Friedman test, I am referring to the user written command -friedman-.
In the literature I have read that the Dunn-Bonferroni test (in Stata implemented by user written command -dunntest-) is the appropriate post-hoc test for the Friedman test. However, I am unsure whether this test accounts for or lets the user account for repeated measurements the same way the Friedman test does. When studying the documentation on -dunntest-, I could not find any indication that this is the case.
A brief elaboration on my research: I have 30 subjects that are measured four times, where each measurement point is several months apart. I want to test whether there is a significant change in the measurements on subject-level across time.
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input int ID double(m1 m2 m3 m4) 4 53.850738525390625 55.07749652862549 39.79530906677246 95.73909759521484 8 162.08384704589844 406.9361877441406 54.08695030212402 173.94705200195313 24 238.5708465576172 135.21643829345703 64.7613353729248 232.27853393554688 29 1376.376220703125 1391.450927734375 1693.4276123046875 881.0337829589844 39 74.0460433959961 112.82498168945313 94.69554901123047 173.03167533874512 43 138.18854522705078 170.98887634277344 153.36073303222656 25.744443893432617 62 356.9570007324219 239.99945068359375 134.4717788696289 232.99710845947266 63 82.06736183166504 203.57208251953125 239.47604370117188 183.60579681396484 64 103.50431823730469 125.83489990234375 192.89240264892578 93.37303161621094 65 168.73209381103516 178.03433227539063 221.94337463378906 203.73833465576172 66 157.960307598114 463.9622802734375 343.06390380859375 79.6368932723999 92 590.0118408203125 109.22186279296875 220.43071746826172 26.350412368774414 103 422.5792541503906 350.69212341308594 331.5404968261719 76.05973052978516 115 660.7237548828125 711.9779968261719 290.68446350097656 286.88926696777344 116 356.1711120605469 330.7752685546875 168.86251831054688 162.41477966308594 137 509.5492248535156 129.39233016967773 436.73602294921875 456.9159851074219 139 71.01797485351563 82.09804916381836 178.60567474365234 138.33961486816406 142 644.3618011474609 1074.0617980957031 759.8912353515625 73.2059326171875 144 90.68435668945313 161.71163940429688 45.937503814697266 70.71932220458984 176 964.295654296875 356.97882080078125 371.00975036621094 182.01382446289063 199 258.25469970703125 672.3479919433594 898.4370727539063 152.0696258544922 203 455.1990661621094 327.36590576171875 124.76653289794922 81.84178161621094 204 641.9442138671875 423.8939971923828 1275.0164184570313 381.62115478515625 205 28.08449363708496 474.22908782958984 648.0500793457031 1029.1197509765625 207 82.31562042236328 142.22564315795898 64.92278289794922 167.06236267089844 208 271.02222442626953 365.1441345214844 342.0950927734375 29.683542251586914 211 609.6959533691406 289.8551025390625 177.98355102539063 241.85422134399414 213 162.07401657104492 233.35426330566406 48.047428131103516 109.10914993286133 217 186.35044860839844 369.10675048828125 947.0944213867188 93.8316764831543 219 429.30731201171875 234.45760345458984 155.66828155517578 134.38890075683594 220 139.77137756347656 427.6331787109375 32.00286865234375 514.7598266601563 end
My data does not follow a normal distribution and I want to account for the fact that the same subjects are measured repeatedly (i.e. that I have dependent groups). Hence, I conducted the Friedman test to test for differences between groups (i.e., the measurement times m1 to m4). The Friedman test was significant.
Code:
friedman m1 m2 m3 m4 Friedman = 59.7702 Kendall = 0.4981 P-value = 0.0010
Thank you very much.
Best,
Tobias
Comment