The neural basis of swap errors in working memory

This is processed data underlying many of the main and supplementary figures in Alleman et al. (2023) https://doi.org/10.1101/2023.10.09.561584It includes time courses of neural activity around the key events of both the retrospective and prospective tasks, as well as relevant metadata from each trial (including the likelihood that the trial falls into each of the discussed response types). The neural activity is filtered with a 500ms width box car filter.To load and use these data, you can refer to the corresponding code, which is available in this github repository: https://github.com/wj2/swap_errorsIn brief, each pickle file contains the data from a particular task in a particular experimental session, and the neural data is binned around a specific time point -- all of this information is given in the file name: lmtc_(task)_all_(time)_(session number).pkl. Each pickle file is a dictionary with the following keys:"spks" : the binned spiketrains across neurons for every trial in the pickle"uc" : the upper color shown on each trial"lc" : the lower color shown on each trial"ps": the posterior probabilities of each response type, derived from the behavioral model. The first column is the probability of a correct response, the second column is a swap response, and the final column is a guess response."cues" : the cue shown on each trial -- 1 means the upper color is the target, 0 means the lower color is the target."other" : a dictionary with a little bit of additional information, most importantly "xs" which is the center of each of the 500 ms bins used for the spikes. Time zero is the time given in the file name (as in the other key in the "other" dictionary, "sequence").