Encoding and Variance Partitioning Analysis

Code for single electrode encoding and variance partitioning analysis. Stimuli were designed to independently vary intonation, phonetics, and speaker. Encoding analyses are equivalent to three-way ANOVA with intonation, sentence, and speaker as the three factors. Each of these categorical variables were coded as dichotomous variables using the simple coding method where regression weights can be interpreted as difference from a reference level of each categorical variable and the intercept is equal to grand mean over all trials.

intonatang.intonation_encoding.dummy_code(sns, sts, speakers, kind='simple')

Used to code categorical variables of sentence, intonation, and speaker.

Simple coding means each condition is compared to a reference level with the intercept being the grand mean. Here, the reference levels are the Neutral condition (st == 1) for intonation, sentence 1 (sn == 1) for sentence, and speaker 3 (sp == 3) for speaker.

The total number of independent variables is 48:
intercept 1 sentence (sn) 4-1 = 3 intonation (st) 4-1 = 3 speaker (sp) 3-1 = 2 sn x st 3*3 = 9 sn x sp 3*2 = 6 st x sp 3*2 = 6 sn x st x sp 3*3*2 = 18
Parameters:
  • sns (list) – list of sentence conditions (1, 2, 3, 4)
  • sts (list) – list of intonation conditions (1, 2, 3, 4)
  • sps (list) – list of speaker conditions (1, 2, 3)
Returns:

  • x with dimensions (n_trials x 48)
Return type:

(ndarray)
intonatang.intonation_encoding.dummy_code_control(sns, sts, speakers, kind='simple')

Used to code categorical variables in non-speech control task

The total number of independent variables here is 40:
intercept 1 sentence (sn) 5-1 = 4 intonation (st) 4-1 = 3 speaker (sp) 2-1 = 1 sn x st 4*3 = 12 sn x sp 4*1 = 4 st x sp 3*1 = 3 sn x st x sp 4*3*2 = 12
Parameters:
  • sns (list) – list of sentence conditions (1, 2, 3, 4, 5)
  • sts (list) – list of intonation conditions (1, 2, 3, 4)
  • sps (list) – list of speaker conditions (1, 2)
Returns:

  • x with dimensions (n_trials x 40)
Return type:

(ndarray)
intonatang.intonation_encoding.dummy_code_varpart(sns, sts, speakers, to_exclude=None)

Returns subsets of coded variables for variance partitioning analysis

to_exclude can be one of the seven following strings: [“sn”, “st”, “sp”, “sn st”, “sn sp”, “st sp”, “sn st sp”]

“sn”, “st”, and “sp” each exclude the variables coding for the main effect of sentence, intonation, and speaker, respectively.

“sn st”, “sn sp”, “st sp”, and “sn st sp” each exclude the variables coding for each pairwise and the three-way interaction.

This function can’t be used to exclude more than one group of predictor variables.

Parameters:to_exclude (str) – which predictor group to exclude
intonatang.intonation_encoding.dummy_code_varpart_control(sns, sts, speakers, to_exclude=None)

Returns subsets of coded variables for variance partitioning analysis for non-speech control

See docstring/documentation for dummy_code_varpart for more information.

Parameters:to_exclude (str) – which predictor group to exclude
intonatang.intonation_encoding.encoding_varpart_permutation_test(Y_mat, sns, sts, sps, n_perms=250, which_chans=array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255]), use_adj_r2=True, control_stim=False)

Runs a permutation test on variance partitioning analysis by shuffling trials. Uses single_electrode_encoding_varpart.

intonatang.intonation_encoding.get_xs_dummy_code_varpart(sns, sts, speakers, control_stim=False)

Returns list of xs for variance partitioning analysis.

Each item in xs contains predictors variables that exclude one group of predictor variables, except the last item in xs which contains all predictor variables (full model with main effects, all pairwise, and three-way interaction)

xs = [x_wo_sn, x_wo_st, x_wo_sp, x_wo_sn_st, x_wo_sn_sp, x_wo_st_sp, x_wo_sn_st_sp, x_all]

intonatang.intonation_encoding.single_electrode_encoding(Y_mat, xs, which_chans=array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255]), use_adj_r2=True, return_weights=False)

Y_mat: n_chans x n_timepoints x n_trials.

intonatang.intonation_encoding.single_electrode_encoding_all_weights(Y_mat, sns, sts, speakers, which_chans=array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255]), control_stim=False)

Returns weights for encoding when using all groups of predictors

intonatang.intonation_encoding.single_electrode_encoding_varpart(Y_mat, sns, sts, sps, which_chans=array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 253, 254, 255]), use_adj_r2=True, control_stim=False)

Returns unique variance of each group of predictors, must use xs from get_xs_dummy_code_varpart

Calculates difference in explained variance between full model and model excluding one group of predictors. The groups of predictors are (sn, st, sp, snxst, snxsp, stxsp, snxstxsp), where x indicates interaction terms. The p_values returned take into account the number of variables in each group.

Parameters:
  • Y_mat (ndarray) – dimensions are n_chans x n_timepoints x n_trials
  • sns (list) – list of sentence conditions (n_trials length)
  • sts (list) – list of intonation conditions (n_trials length)
  • sps (list) – list of speaker conditions (n_trials length)
  • which_chans (list) – list of channels to do encoding analysis on
  • use_adj_r2 – use adjusted r2 instead of r2
  • control_stim – whether analysis is on non-speech control task (changes coding of categorical variables)
Returns:

  • r2s_varpart (ndarray): dimensions n_chans x n_timepoints x 7 (groups of predictors)
  • p_values (ndarray): dimensions n_chans x n_timepoints x 7
  • f_stats (ndarray): dimensions n_chans x n_timepoints x 7
Return type:

(tuple)