If I have a df like this:
normalized_0 normalized_1 normalized_0 mean std
Site
0 NaN 0.798262 1.456576 0.888687 0.118194
1 0.705540 0.885226 NaN 0.761488 0.047023
2 0.669539 1.002526 1.212976 0.826657 0.077940
3 0.829826 0.968180 0.988679 0.871290 0.032367
How do I calculate a two sided t-test for 0, 1, 2 vs. 3?
I tried it with:
from scipy.stats import ttest_ind
df['ttest'] = ttest_ind(df, d.loc[3])
But this does not work... The error I get is:
TypeError: unsupported operand type(s) for /: 'str' and 'int'
How would you solve this?
My answer might be completely off, as I've only read about t-tests :)
What I understood from your question is that you have a table with both normalized values and their descriptive statistics (mean, std).
Each index value within this table is a category of your analysis, and you want to compare categories [0, 1, 2] vs [3].
I also assume you only need normalized values as input arrays, without mean or std.
selected_data = df.copy()
selected_data = selected_data[['normalized_0', 'normalized_1', 'normalized_0.1']]
selected_data['ttest'] = [ttest_ind(a=selected_data.iloc[3, :].values, \
b=selected_data.iloc[x, :].values, \
nan_policy='omit') for x in np.arange(len(selected_data))]
df.join(selected_data['ttest'])
normalized_0 normalized_1 normalized_0.1 mean std ttest
Site
0 NaN 0.798262 1.456576 0.888687 0.118194 (-0.7826642930343911, 0.4909212050511221)
1 0.705540 0.885226 NaN 0.761488 0.047023 (1.4370158341444121, 0.24625840339538163)
2 0.669539 1.002526 1.212976 0.826657 0.077940 (-0.19764518466194855, 0.8529602343240825)
3 0.829826 0.968180 0.988679 0.871290 0.032367 (0.0, 1.0)
a and b parameters are row values of selected columns
# values of third category for example
selected_data.iloc[3, :].values
# array([0.829826, 0.96818 , 0.988679])
omit is to ignore nan values when calculating test (by default the parameter of nan_policy is set to propagate which returns nan if any missing values are present).