Figure 3 - Source Data 1

Cross-validated model likelihood during stable period for example mice (each data is the median value of corresponding likelihood distribution, each row contains data from one animal)

Model 1 (full model)	Model 2 (1 = 0)	Model 3 (2 = 0)	Model 4 (1 = 0, 2 = 0)	Model 5 ( = 0)	Model 6 (1 = 1)	Model 7 (2 = 0)	RL Model
0.637718238070385	0.681494222740181	0.672743363467017	0.683623822796313	0.677460300908706	0.558369197666465	0.643107043222649	0.679313784355047
0.66494018566057	0.668034854415511	0.664915996714429	0.66688776009153	0.664702102753042	0.560322961150174	0.667166045943659	0.664188270610011
0.649891758094156	0.649060928976284	0.647276181185805	0.649465425391352	0.6520713387144	0.576371315512116	0.652509875429021	0.65136884042475
0.678422622728384	0.675392228617203	0.67003964268349	0.675580927521027	0.679645517234712	0.558471064186295	0.678494385935778	0.678286813211368
0.71871031679855	0.708868393876576	0.736673409495222	0.736280460963109	0.734766385915832	0.560685112325594	0.731764987147554	0.724960944360982
0.709248332269017	0.70210878913042	0.712198705686216	0.715695210476829	0.714337205306857	0.56122653191751	0.712985056150687	0.710699410530242
0.729930017462512	0.739834313951927	0.74300001415788	0.748847907877292	0.748666737026045	0.556401406309569	0.745395278101237	0.756232090836685
0.697619799205088	0.701731997147063	0.701429827061449	0.701920002130815	0.692738760617534	0.584957027140611	0.698400321594835	0.700139624117447
0.744938152902238	0.738368065626469	0.744369866070023	0.744665531041711	0.738404169375007	0.56471945171425	0.733650934123624	0.73885121173563
0.694746218224772	0.713468660148001	0.719513043524408	0.717284717788781	0.689261460030278	0.592747822047497	0.698050832527689	0.716367514528272
___________________________________
Test results for Model 2 vs Model 1


Column B	Model 2 (1 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.5566
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	34 , -21
  Sum of signed ranks (W)	13
  Number of pairs	10
	
Median of differences	
  Median	0.001132
	
How effective was the pairing?	
  rs (Spearman)	0.8424
  P value (one tailed)	0.0019
  P value summary	**
  Was the pairing significantly effective?	Yes

___________________________________
Test results for Model 3 vs Model 1

Column C	Model 3 (2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.1309
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	43 , -12
  Sum of signed ranks (W)	31
  Number of pairs	10
	
Median of differences	
  Median	0.00338
	
How effective was the pairing?	
  rs (Spearman)	0.8909
  P value (one tailed)	0.0006
  P value summary	***
  Was the pairing significantly effective?	Yes

___________________________________
Test results for Model 4 vs Model 1

Column D	Model 4 (1 = 0, 2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0371
  Exact or approximate P value?	Exact
  P value summary	*
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	48 , -7
  Sum of signed ranks (W)	41
  Number of pairs	10
	
Median of differences	
  Median	0.005374
	
How effective was the pairing?	
  rs (Spearman)	0.8788
  P value (one tailed)	0.0008
  P value summary	***
  Was the pairing significantly effective?	Yes


___________________________________
Test results for Model 5 vs Model 1

Column E	Model 5 ( = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.3750
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	37 , -18
  Sum of signed ranks (W)	19
  Number of pairs	10
	
Median of differences	
  Median	0.001701
	
How effective was the pairing?	
  rs (Spearman)	0.9515
  P value (one tailed)	<0.0001
  P value summary	****
  Was the pairing significantly effective?	Yes

___________________________________
Test results for Model 6 vs Model 1

Column F	Model 6 (1 = 1)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0020
  Exact or approximate P value?	Exact
  P value summary	**
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	0 , -55
  Sum of signed ranks (W)	-55
  Number of pairs	10
	
Median of differences	
  Median	-0.1163
	
How effective was the pairing?	
  rs (Spearman)	0.05455
  P value (one tailed)	0.4458
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for Model 7 vs Model 1

Column G	Model 7 (2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0488
  Exact or approximate P value?	Exact
  P value summary	*
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	47 , -8
  Sum of signed ranks (W)	39
  Number of pairs	10
	
Median of differences	
  Median	0.002961
	
How effective was the pairing?	
  rs (Spearman)	0.9879
  P value (one tailed)	<0.0001
  P value summary	****
  Was the pairing significantly effective?	Yes

___________________________________
Test results for RL Model vs Model 1

Column H	RL Model
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0645
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	46 , -9
  Sum of signed ranks (W)	37
  Number of pairs	10
	
Median of differences	
  Median	0.001998
	
How effective was the pairing?	
  rs (Spearman)	0.8788
  P value (one tailed)	0.0008
  P value summary	***
  Was the pairing significantly effective?	Yes

___________________________________
Test results for RL Model vs Model 4

Column H	RL Model
vs.	vs.
Column D	Model 4 (1 = 0, 2 = 0)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.3223
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	17 , -38
  Sum of signed ranks (W)	-21
  Number of pairs	10
	
Median of differences	
  Median	-0.00224
	
How effective was the pairing?	
  rs (Spearman)	1
  P value (one tailed)	<0.0001
  P value summary	****
  Was the pairing significantly effective?	Yes

_______________________________________________________________________________________________________________________________________________________________________
Cross-validated model likelihood during switching period for example mice (each data is the median value of corresponding likelihood distribution, each row contains data from one animal)

Model 1 (full model)	Model 2 (1 = 0)	Model 3 (2 = 0)	Model 4 (1 = 0, 2 = 0)	Model 5 ( = 0)	Model 6 (1 = 1)	Model 7 (2 = 0)	RL Model
0.659545453739007	0.657419673430703	0.630164901614051	0.654554076658644	0.629736968833532	0.546258812950634	0.581818260684444	0.642387168191371
0.642246998155843	0.644025093592386	0.638528579703755	0.641649719762962	0.599718099271134	0.565079436234697	0.642928057153889	0.641606677560601
0.630301137726123	0.636697910856166	0.641523237897364	0.640321159744048	0.616468935393803	0.53609928459196	0.611153947368635	0.632529724946681
0.635650692560628	0.643782946982389	0.637135257063893	0.642810556065457	0.642244062889	0.504292713728132	0.589758472084201	0.638382240054696
0.661826207559436	0.663553396232213	0.664588029228341	0.667937983794976	0.563965361295274	0.570729940692991	0.565850869347407	0.647400104187471
0.621080209558828	0.637392810389452	0.647830682267915	0.652158451802411	0.584023599132878	0.590172609654828	0.598710755742129	0.648143409680543
0.655079004262679	0.658022476991836	0.65934288716491	0.662618974528468	0.636573409018783	0.527542641783524	0.638413013759701	0.653153336699021
0.652762860246368	0.668411889428061	0.628314835586068	0.659616959732197	0.622094866301896	0.586701385141525	0.615679313562565	0.646530079400125
0.666764040705305	0.669149202729819	0.667762535819877	0.667735540653424	0.664605321382306	0.53022976686289	0.64016609083253	0.667490198905784
0.648457044332429	0.649837229101203	0.647615133402028	0.65187626202887	0.578088478779527	0.567640278640158	0.648134585439447	0.64574040132273

___________________________________
Test results for Model 2 vs Model 1

Column B	Model 2 (1 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0137
  Exact or approximate P value?	Exact
  P value summary	*
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	51 , -4
  Sum of signed ranks (W)	47
  Number of pairs	10
	
Median of differences	
  Median	0.002664
	
How effective was the pairing?	
  rs (Spearman)	0.903
  P value (one tailed)	0.0004
  P value summary	***
  Was the pairing significantly effective?	Yes

___________________________________
Test results for Model 3 vs Model 1

Column C	Model 3 (2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.7695
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	31 , -24
  Sum of signed ranks (W)	7
  Number of pairs	10
	
Median of differences	
  Median	0.001242
	
How effective was the pairing?	
  rs (Spearman)	0.3455
  P value (one tailed)	0.1652
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for Model 4 vs Model 1

Column D	Model 4 (1 = 0, 2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0195
  Exact or approximate P value?	Exact
  P value summary	*
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	50 , -5
  Sum of signed ranks (W)	45
  Number of pairs	10
	
Median of differences	
  Median	0.006483
	
How effective was the pairing?	
  rs (Spearman)	0.8182
  P value (one tailed)	0.0029
  P value summary	**
  Was the pairing significantly effective?	Yes

___________________________________
Test results for Model 5 vs Model 1

Column E	Model 5 ( = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0059
  Exact or approximate P value?	Exact
  P value summary	**
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	2 , -53
  Sum of signed ranks (W)	-51
  Number of pairs	10
	
Median of differences	
  Median	-0.03024
	
How effective was the pairing?	
  rs (Spearman)	0.2485
  P value (one tailed)	0.2459
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for Model 6 vs Model 1

Column F	Model 6 (1 = 1)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0020
  Exact or approximate P value?	Exact
  P value summary	**
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	0 , -55
  Sum of signed ranks (W)	-55
  Number of pairs	10
	
Median of differences	
  Median	-0.09265
	
How effective was the pairing?	
  rs (Spearman)	-0.1515
  P value (one tailed)	0.3410
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for Model 7 vs Model 1

Column G	Model 7 (2 = 0)
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0059
  Exact or approximate P value?	Exact
  P value summary	**
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	2 , -53
  Sum of signed ranks (W)	-51
  Number of pairs	10
	
Median of differences	
  Median	-0.02448
	
How effective was the pairing?	
  rs (Spearman)	-0.04242
  P value (one tailed)	0.4592
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for RL Model vs Model 1

Column H	RL Model
vs.	vs.
Column A	Model 1 (full model)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.6250
  Exact or approximate P value?	Exact
  P value summary	ns
  Significantly different (P < 0.05)?	No
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	22 , -33
  Sum of signed ranks (W)	-11
  Number of pairs	10
	
Median of differences	
  Median	-0.001283
	
How effective was the pairing?	
  rs (Spearman)	0.5394
  P value (one tailed)	0.0569
  P value summary	ns
  Was the pairing significantly effective?	No

___________________________________
Test results for RL Model vs Model 4

Column H	RL Model
vs.	vs.
Column D	Model 4 (1 = 0, 2 = 0)
	
Wilcoxon matched-pairs signed rank test	
  P value	0.0020
  Exact or approximate P value?	Exact
  P value summary	**
  Significantly different (P < 0.05)?	Yes
  One- or two-tailed P value?	Two-tailed
  Sum of positive, negative ranks	0 , -55
  Sum of signed ranks (W)	-55
  Number of pairs	10
	
Median of differences	
  Median	-0.006964
	
How effective was the pairing?	
  rs (Spearman)	0.8303
  P value (one tailed)	0.0024
  P value summary	**
  Was the pairing significantly effective?	Yes

