import glob import os import pandas as pd import matplotlib.pyplot as plt import time import numpy as np import seaborn as sns import math path = os.getcwd() def get_average(records): return sum(records) / len(records) def get_variance(records): average = get_average(records) return sum([(x - average) ** 2 for x in records]) / len(records) def get_standard_deviation(records): variance = get_variance(records) return math.sqrt(variance) def get_rms(records): return math.sqrt(sum([x ** 2 for x in records]) / len(records)) def get_mse(records_real, records_predict): if len(records_real) == len(records_predict): return sum([(x - y) ** 2 for x, y in zip(records_real, records_predict)]) / len(records_real) else: return None def get_rmse(records_real, records_predict): mse = get_mse(records_real, records_predict) if mse: return math.sqrt(mse) else: return None def get_mae(records_real, records_predict): if len(records_real) == len(records_predict): return sum([abs(x - y) for x, y in zip(records_real, records_predict)]) / len(records_real) else: return None def writeSDCSV(filename): file = pd.read_csv("Mean.csv") conditions = file['condition'] dict = {} dict['conditon'] = conditions for scale in scales: temp = [] for condition in conditions: col = df_merged.groupby('condition').get_group(condition) col = col[scale] temp.append(get_standard_deviation(col)) dict[scale] = temp df = pd.DataFrame(dict) df.to_csv(filename) def draw(scale): conditions = file['condition'] result = file[scale] plt.figure(figsize=(9, 6), dpi=100) sd = pd.read_csv(SD) std_err = sd[scale] error_params=dict(elinewidth=1,ecolor='black',capsize=5) plt.bar(conditions, result, width=0.35, color=colors,alpha=a,yerr=std_err,error_kw=error_params) plt.title(scale,fontsize=15) plt.ylabel('score') plt.grid(alpha=0, linestyle=':') plt.savefig(scale, dpi=300) #plt.show() def drawTogether(): scales = ["mental-demand","physical-demand","temporal-demand","performance", "effort","frustration","total"] scales = ["total","mental-demand","physical-demand","frustration"] plt.figure(figsize=(15,7)) x = np.arange(len(scales)) total_width, n = 0.8, 4 width = total_width / n for i in range(0,4): result = [] std_err = [] sd = pd.read_csv(SD) for scale in scales: result.append(file.iloc[i][scale]) std_err.append(sd.iloc[i][scale]) error_params=dict(elinewidth=1,ecolor='black',capsize=5) plt.bar(x+width*(i-1),result,width=width,color=colors[i],label=file.iloc[i]["condition"],alpha=a,yerr=std_err,error_kw=error_params) plt.legend() # plt.title("TLX Average",fontsize=15) plt.xticks(x+width/2,scales) #plt.show() plt.ylabel('score') plt.savefig("summary.jpg",dpi=300) # Merge all the .csv file start with "HectorVR", and all_files = glob.glob(os.path.join(path, "HectorVR*.csv")) df_from_each_file = (pd.read_csv(f, sep=',') for f in all_files) df_merged = pd.concat(df_from_each_file, ignore_index=True) # Save the file to Merged.csv in the same folder df_merged.to_csv( "Merged.csv") # save the results in csv file = df_merged.groupby(["condition"]).mean() file.to_csv("Mean.csv") scales = ["mental-demand","physical-demand","temporal-demand","performance", "effort","frustration","total"] SD = "standard_deviation.csv" writeSDCSV(SD) file = pd.read_csv("Mean.csv") colors = sns.color_palette() a = 0.6 for scale in scales: draw(scale) drawTogether()