Please enable JavaScript.

Coggle requires JavaScript to display documents.

Python (trianhnguyen), Thống kê cơ bản - Coggle Diagram

- - - - for-list
        house = [["hall", 11.25],
        ["kit", 18.0],
        ["livin", 20.0],
        ["bath", 9.50]]
        for h,m in house:
        if h=="kit":
        continue
        if h=="livin":
        break
        print("the "+str(h)+" is "+str(m)+" sqm")
      - for-dict
        for key, val in my_dict.items():
      - for-numpy array
        for val in np.nditer(my_array)
      - for-dataframe
        for lab, row in my_dict.iterrows() :
        print(lab+": "+str(row["cars_per_cap"]))
  - - - def my_function(fname, lname):
        print(fname + " " + lname)
      - Arbitrary Arguments, *args
        def my_function(*kids):
        print("The youngest child is " + kids[2])
        my_function("Emil", "Tobias", "Linus")
      - Keyword Arguments
        def my_function(child3, child2, child1):
        print("The youngest child is " + child3)
        my_function(child1 = "Emil", child2 = "Tobias", child3 = "Linus")
      - Arbitrary Keyword Arguments, **kwargs
        def my_function(**kid):
        print("His last name is " + kid["lname"])
        my_function(fname = "Tobias", lname = "Refsnes")
  - - - dict = {'name': 'john','code':6734, 'dept': 'sales'}
    - - list = [ 'abcd', 786 , 2.23, 'john', 70.2 ]
        mutable
      - Tuple
        immutable
        
        tuple = ( 'abcd', 786 , 2.23, 'john', 70.2 )
        
        convert
        tuple(s)
      - range
    - - str = 'Hello World!'
      - convert
        str(x)
    - - var1 = 1
      - convert
        int(x [,base])
        long(x [,base] )
        float(x)
        complex(real [,imag])
    - - myset = {"apple", "banana", "cherry"}
- - - - file = open(filename, mode='w')
        print(file.read())
        file.close()
      - with open("abc.txt","r") as file:
        print(file.readline())
      - import float/string dtype
        data= np.loadtxt(filename, delimiter="\t",dtype="float")
        data= np.loadtxt(filename, delimiter=",", skiprows=1, usecols=[0,2])
        mix dtype
        data = np.genfromtxt(filename, delimiter=',', names=True, dtype=None)
        csv
        data = np.recfromcsv(filename, delimiter=",", names=True, dtype=None)
      - import pickle
        Open pickle file and load data: d
        with open('data.pkl', 'rb') as file:
        d = pickle.load(file)
      - load excel file
        xls = pd.ExcelFile(file)
        df2 = xls.parse(0, usecols=[0], skiprows=[0], names=['Country'])
      - from sas7bdat import SAS7BDAT
        with SAS7BDAT('sales.sas7bdat') as file:
        df_sas = file.to_data_frame()
      - Stata file
        df= pd.read_stata('disarea.dta')
- - - - merge
        
        single table
        newdf = df1.merge(df2, on="col1", suffixes("_col1","_col2", how="left", indicator=False, validate="one_to_one")
        newdf = df1.merge(df2, on=["col1","col2"])
        how
        how:‘left’, ‘right’, ‘outer’, ‘inner’, ‘cross’, default=‘inner’
        index table
        newdf = df1.merge(df2, how="right", left_on="col1", right_on="col2", left_index=True, right_index=True)
        
        merge_ordered
        pd.merge_ordered(df1,df2, on="col1", suffixes=("_df1","_df2")
        ,how="outer"
        ,fill_method="ffill")
        
        pd.merge_asof(df1,df2, on='col1', suffixes=('_df1','_df2')
        ,direction='nearest')
      - concat df
        
        concat
        pd.concat([table1, table2], ignore_index=False, key=['table1','table2'], sort=True, join='inner', verify_integrity=True)
        .append
        table1.append([table2], ignore_index=True, sort=True)
  - - - sum
        df["col1"].sum()
        df["col1"].cumsum()
      - min-max
        df["col1"].min()
        df["col1"].max()
        df["col1"].cummax()
        df["col1"].cummin()
      - count
        df["col1"].value_counts(sort=True)
        df["col1"].value_counts(normalize=True)
        df['col1'].count()
      - df["col1"].cumprod()
      - to apply a function or a list of function names
        df.agg([func1,func2])
        .agg([np.min,np.max])
    - - mean
        np.mean(df["col1"])
        df["col1"].mean()
      - standard deviation
        df["col1"].std()
      - median
        df["col1"].median()
      - Correlation
        df.corr()
        np.corrcoef(x,y)
        df['col1'].corr(df['col2'])
        0.99: very strong| 0.75: strong| 0.56: moderate| 0.21: weak| 0.04: no relationship
      - mode
        df["col1"].mode()
        scipy.stats.mode(df['col1'])
      - iqr
        scipy.stats.iqr(df['col1'])
        df["col1"].quantile(q=[0.25,0.5])
      - var
        df["col1"].var()
    - - The binomial distribution
        
        binom.rvs(# coin, Prob of success, size=# N trial)
        binom.rvs(1, 0.5,5)
        binom.pmf(num heads, num trials, prob of heads)
        binom.pmf(7,10,0.5)
        binom.cdf(7,10,0.5)
      - Discrete distributions
        
        np.mean(df.sample(10, replace=True)
      - Continuous distributions
        
        uniform.cdf(x1,min, max)
        rvs == linespace
        uniform.rvs(0, 5, size=10)
      - Normal distribution
        
        normal distribution
        norm.cdf(x, mean, std)
        norm.ppf(x, ,mean, std)
        norm.rvs(mean,std, size=size)
      - Poisson
        
        poisson.pmf(k, mu, loc=0)
        poisson.cdf(k, mu, loc=0)
        rvs(mu, loc=0, size=1, random_state=None)
      - exponential distribution
        expond.rvs(loc=0, scale=1, size=1, random_state=None)
        expon.cdf(x, loc=0, scale=1)
        expon.pdf(x, loc=0, scale=1)
      - other distribution
        t-distribution
        
        log-normal distribution
    - - my_house = np.array([18.0, 20.0, 10.75, 9.50])
        [18. 20. 10.75 9.5 ]
      - Compare
        <,>,==,!=,>=,<=
        my_house>=18
        [ True True False False]
        np.logical_and(my_house>10,my_house<18) np.logical_or(my_house>10,my_house<18) np.logical_not(my_house>10,my_house<18)
      - random
        np.random.seed(123)
        np.random.rand()
        np.random.randint(1,7)
        df.sample(2,replacce=True)
      - numpy numpy
        
        array
        
        np.array(python_list, dtype=np.float32)
        np.zeros((row, col))
        np.random.random((row,col))
        np.arange(start_value, end_value)
        
        attribute
        .shape
        .dtype
        method
        .flatten()
        .reshape(row, col)
        .astype(np.int32)
        
        index-slicing
        row
        nparr[row]
        col
        nparr[:,col]
        nparry[col1:col3)
        nparry[row1:row3:step, col1:col3:step)
        
        sort
        np.sort(nparr, axis=0)
        
        fancy indexing
        nparr[nparr[:,1]%2==0]
        
        np.where(nparr[:,1]%2==0)
        
        find and replace
        np.where(nparr==0, "replacevalue", nparr)
        
        concat
        np.concatenate((nparr1, nparr2))
        np.concatenate((nparr1, nparr2), axis=1)
        
        np.delete(nparr,col, axis=1)
        
        flip/transpose
        np.flip(nparr, axis=(0,1))
        np.transpose(nparr, axes=(1,0,2))
        stack/split
        np.split(nparr,3, axis=2)
        np.stack([nparr1, nparr2], axis=2)
        
        nparr aggregate func
        .sum()
        .sum(axis=0/1, keepdims=True/False)
        .min()
        .max()
        .mean()
        .cumsum()
        
        nparr + 3
        nparr *2
        nparr /2
        nparr >2
        vectorize function
        np.vectorize(len)(nparr)
        vectorize method
        np.vectorize(str.upper)(nparr)
        
        boarcasting
        Two dimensions are compatible to * / + - when
        they are equal, or
        one of them is 1
        
        load .npy file
        with open('file.npy','rb') as f:
        rgb_array=np.load(f)
        save:
        with open('flie.npy','wb') as f:
        np.save(f, nparr)
- - - - Pandas plot
        
        bar chart
        df.plot(kind="bar", title="title")
        plt.show()
        
        line chart
        df.plot(x="col1", y="col2",kind="line",rot=45)
        plt.show()
        
        scatter chart
        df.plot(x="col1", y="col2",kind="scatter")
        plt.show()
        
        histogram
        df.hist()
        df[["col1", "col2"]].hist()
        plt.legend(["legend1","legend2"])
        plt.show()
      - Matplotlib
        from matplotlib import pyplot as plt
        
        line
        
        plt.plot(df.times, df.distance,label="region",color="blue", linewidth=1,linestyle='--',marker='x')
        plt.xlabel("times")
        plt.ylabel("distance")
        plt.title("times-distance", fontsize=12)
        plt.style.use('fivethirtyeight')
        plt.text(15, 120, "this is linear")
        plt.show()
        
        scatter plot
        
        plt.scatter(df.age, df.income,color='green', marker='s',alpha=0.9)
        plt.xlabel("age")
        plt.ylabel("income")
        plt.title("age-income", fontsize=12)
        plt.style.use('fivethirtyeight')
        plt.text(30, 590, "this is not linear")
        plt.show()
        
        plt.bar(df.precinct, df.pets_abducted)
        
        plt.barh(df.precinct, df.pets_abducted)
        
        plt.bar(df.precinct, df.pet_abductions,
        yerr=df.error)
        
        Stacked bar
        
        plt.bar(df.precinct, df.dog,label='Dog')
        plt.bar(df.precinct, df.cat, bottom=df.dog, label='Cat')
        
        Histograms
        
        plt.hist(gravel.mass,bins=40,range=(50, 100), density=False)
        
        matplotlib
        import matplotlib.pyplot as plt
        
        default fig
        fig, ax = plt.subplots()
        create fig 2rx1c
        fig, ax = plt.subplots(2,1,sharey=True)
        
        plot df
        ax.plot(df['col1'], df['col2'], color='b',marker='o',linestyle='--')
        ax[0].plot(df['col1'], df['col2'])
        
        Customize the x-axis label
        ax.set_xlabel("Label")
        Customize the y-axis label
        ax.set_ylabel("Label")
        Add the title
        ax.set_title("Title")
        annotate
        ax.annotate('>1 degree', (pd.Timestamp('2015-10-06'), 1))
        ax2.annotate(">1 degree", xy=(pd.Timestamp('2015-10-06'),1), xytext=(pd.Timestamp('2008-10-06'),-0.2), arrowprops={"arrowstyle":"->", "color":"gray"})
        legen
        ax.legend()
        tick label
        ax.set_ticklabels(df.index, rotation=90)
        
        add twince ax
        
        ax2= ax.twinx()
        ax2.plot(df['col1'], df['col2'], color='red')
        ax.set_ylabel("Label")
        ax.tick_params('y', colors='red')
        
        Define a function called plot_timeseries
        def plot_timeseries(axes, x, y, color, xlabel, ylabel):
        axes.plot(x, y, color=color)
        axes.set_xlabel(xlabel)
        axes.set_ylabel(ylabel, color=color)
        axes.tick_params('y', colors=color)
        
        bar
        ax.bar(df.index, df['col1'])
        stack bar chart
        ax.bar(df.index, df['col1'], bottom=df['col2'])
        add error
        ax.bar('label1', df['col1'].mean(), yerr=df.['col1'].std())
        
        Histograms
        ax.hist(df['col1'], label='label1', bins=5,
        histtype='step')
        
        ax.errorbar(df.index, df['col1'], yrr=df['col3'])
        boxplot
        ax.boxplot([df['col1'], df['col2']])
        ax.set_xticklabels(['label1', 'label2'])
        ax.set_ylabel('y_label')
        
        ax.scatter(df['col1'], df['col2'], color='color', label='label',
        c=df.index)
        
        #### style - themes
        plt.style.use("default")
        plt.style.use("ggplot")
        plt.style.use("seaborn-colorblind")
        
        save fig to file
        fig.set_size_inches([w,h])
        fig.savefig('pic.png', dpi=300)
        fig.savefig('pic.jpg', quality=50)
        fig.savefig('pic.svg')
        
        cheat sheet
        https://datacamp-community-prod.s3.amazonaws.com/e1a8f39d-71ad-4d13-9a6b-618fe1b8c9e9
      - Seaborn
        import matplotlib.pyplot as plt
        import seaborn as sns
        
        scatter
        sns.scatterplot(x='col1', y='col2', data=df,
        hue="col3",
        hue_order=["col3val1","col3val2"])
        or set color by palette
        hue_color={"col3val1":"black","col3val2":"red"}
        hue="col3",
        , palette=hue_color)
        
        countplot
        sns.countplot(x='col1', data=df,
        hue='col2')
        g.set_titles("this is {col_name}")
        
        df = sns.load_dataset("datasetname")
        
        relplot
        
        sns.relplot(x='col1', y='col2', data=df,
        kind='line',
        markers=True,
        dashes=False,
        style="col3",
        hue="col3",
        ci="sd"
        'sd': standard deviation
        
        sns.relplot(x='col1', y='col2', data=df,
        kind='scatter',
        col='col3',
        row='col4',
        col_wrap=2,
        col_order=["val1","val2"],
        row_order==["val1","val2"]
        hue="col3",
        size="col3",
        style="col3",
        alpha=0.9)
        Parameters
        col/row: for subplot
        
        catplot
        
        count
        g= sns.catplot(x="col1", data=df,
        kind="count",
        order=["val1","val2"],
        col='col3',
        row='col4',
        hue='col5')
        parameter
        x: bar
        y: hbar
        title
        g.fig.suptitle('title', y=1,05)
        g.set_titles("this is {col_name}")
        g.set(xlabel="label1", ylabel="label2")
        rot
        plt.xticks(rotation=90)
        
        box
        
        sym= "",
        whis=2.0 or whis = [5,95]
        
        bar/hbar
        sns.catplot(x='col1', y="col2", data=df,
        kind="bar",
        ci=None)
        hbar: x<=>y
        
        point plot
        
        sns.catplot(x='col1', y='col2', data=df,
        kind='point',
        join=Fasle,
        hue='col3',
        capsize=0.2,
        ci=None,
        estimator=np.median)
        
        style
        sns.set_style('stylename')
        palette
        sns.set_palette('palettename')
        customepalette=['red','blue']
        context
        sns.set_context('paper')
        
        sns.lmplot(x='col1', y='col2', data=df, ci=None)
        
        sns.boxplot(x='col1', y='col2', data=df)
        
        sns.displot(df['Award_Amount'])