DATA = "prolifc_data_combine_num_each_pp.xlsx"
DATA2 = "prolifc_data_each_pp.xlsx"
winsize = 0.6
# ANOVA within subject clustering (5) * type (2) for each winsize
data = pd.read_excel(PATH + DATA)
data = data[data["winsize"] == winsize]
aov = pg.rm_anova(data=data,
dv="mean_deviation_score",
within=["percent_triplets", "protectzonetype"],
subject="participant")
posthocs = pg.pairwise_ttests(
dv="mean_deviation_score",
within=["percent_triplets", "protectzonetype"],
subject="participant",
data=data,
padjust="fdr_bh",
effsize="cohen")
# ANOVA within subject
data2 = pd.read_excel(PATH + DATA2)
data2 = data2[data2["winsize"] == 0.4] # winsize 0.4 unblanced data
aov_table = AnovaRM(
data=data2,
depvar="mean_deviation_score",
subject="participant",
within=["protectzonetype", "numerosity", "percent_triplets"]).fit()
aov_table.summary()
# so we create new columns
subs = 3*subjs
subs.sort()
conds = ["N","P","T"] * len(subjs)
laut = []
for i in range(len(subjs)):
laut.append(group_df["N_Laut"][i])
laut.append(group_df["P_Laut"][i])
laut.append(group_df["T_Laut"][i])
ang = []
for i in range(len(subjs)):
ang.append(group_df["N_Ang"][i])
ang.append(group_df["P_Ang"][i])
ang.append(group_df["T_Ang"][i])
# and throw them in a dict to make a dataframe
dict = {"Subject":subs,"Bed":conds,"Laut":laut,"Ang":ang}
temo_long = pd.DataFrame(dict)
# now we can calculate the rmANOVAs for each Laut & Ang (see how we imported AnovaRM from statsmodels above)
anova_laut = AnovaRM(temo_long,depvar="Laut",subject="Subject",within=["Bed"],aggregate_func=None).fit()
print(anova_laut.summary())
tukey_hsd_laut = pairwise_tukeyhsd(temo_long.Laut, groups=temo_long.Bed, alpha=0.05)
print(tukey_hsd_laut.summary())
anova_ang = AnovaRM(temo_long,depvar="Ang",subject="Subject",within=["Bed"],aggregate_func=None).fit()
print(anova_ang.summary())
tukey_hsd_ang = pairwise_tukeyhsd(temo_long.Ang, groups=temo_long.Bed, alpha=0.05)
print(tukey_hsd_ang.summary())
## Example for saving a csv file from dataframe to read with another program (e.g. R or Excel)
# temo_df.to_csv("{}TEMO.csv".format(dir)) # define directory and file name to your need...
sub_id = sub_id + 1
for ind_t, task in enumerate(task_list):
for ind_c, con in enumerate(condition_list):
# generate random value here as example
my_val = np.random.normal(ind_c + ind_t, 1, 1)[0]
df_full.loc[my_row] = [
group, lan, sub_id, task, con, my_val
]
my_row = my_row + 1
# conduct ANOVA using mixedlm
my_model_fit = smf.mixedlm("my_value ~ group * language * task * condition",
df_full,
groups=df_full["sub_id"]).fit()
# get fixed effects
my_model_fit.summary()
# get random effects
my_model_fit.random_effects
# 4-way analysis possible in ols, but random effects not accounted for!
my_model_fit = smf.ols(
formula='my_value ~ group * language * task * condition',
data=df_full).fit()
sm.stats.anova_lm(my_model_fit, typ=2)
my_model_fit.summary()
# --------
# Save dataframes as csv
# --------
df_2way.to_csv('df_2way.csv', index=False)
df_full.to_csv('df_full.csv', index=False)
