def macro_dlog_mixed(endog, exo_vars, sheet, coef, start_period, end_period,lags_endog=None, exo_lags=None, diff_exo_vars=None):
    """
    Forecasts 'endog' using a log-difference regression model with optional differencing
    and lagging of exogenous variables. Exogenous variables are sourced from Assump,
    sheet, and (optionally) Projection Sheets.

    Parameters:
        endog (str): Name of the endogenous variable to forecast.
        exo_vars (list): List of exogenous variable names.
        sheet (DataFrame): DataFrame to store forecasts and read some exogenous data from.
        coef (list): Regression coefficients (intercept + exo + lags).
        start_period (int): Start year of the forecast.
        end_period (int): End year of the forecast.
        lags_endog (list): Lags of endogenous variable to include (in differences).
        exo_lags (dict): Dictionary of lags per exogenous variable.
        diff_exo_vars (list): Exogenous variables that should be differenced.

    Returns:
        None: Forecasts are written in-place to `sheet`.
    """
    if lags_endog is None:
        lags_endog = [1]
    if exo_lags is None:
        exo_lags = {var: [0] for var in exo_vars}
    if diff_exo_vars is None:
        diff_exo_vars = []

    global Assump, AB_Projections

    def get_val(var, t):
        """Get value of a variable at time t from Assump, sheet, or AB_Projections."""
        if var in Assump and t in Assump[var]:
            return Assump[var][t]
        elif var in sheet.index and t in sheet.columns:
            return sheet.loc[var, t]
        elif sheet is CER_Projections and var in AB_Projections.index and t in AB_Projections.columns:
            return AB_Projections.loc[var, t]
        else:
            return None

    for i in range(start_period, end_period + 1):
        if i in sheet.columns and pd.notna(sheet.loc[endog, i]) and sheet.loc[endog, i] != '':
            continue

        X = [1]  # Intercept

        # Exogenous variables
        for var in exo_vars:
            for lag in exo_lags.get(var, [0]):
                t = i - lag
                if var in diff_exo_vars:
                    t_minus_1 = t - 1
                    val_t = get_val(var, t)
                    val_tm1 = get_val(var, t_minus_1)
                    if val_t and val_tm1:
                        dlog_val = np.log(val_t) - np.log(val_tm1)
                    else:
                        dlog_val = 0
                    X.append(dlog_val)
                else:
                    val = get_val(var, t)
                    log_val = np.log(val) if val else 0
                    X.append(log_val)

        # Lagged differenced endogenous variable
        for lag in lags_endog:
            t_lag = i - lag
            if t_lag - 1 in sheet.columns and t_lag in sheet.columns:
                y_t = sheet.loc[endog, t_lag]
                y_tm1 = sheet.loc[endog, t_lag - 1]
                if pd.notna(y_t) and pd.notna(y_tm1) and y_t != '' and y_tm1 != '':
                    dlog_y = np.log(y_t) - np.log(y_tm1)
                else:
                    dlog_y = 0
            else:
                dlog_y = 0
            X.append(dlog_y)

        # Predict Δlog(y)
        dlog_y_t = sum(c * x for c, x in zip(coef, X))

        # Get log(y_{t-1})
        if (i - 1) in sheet.columns and pd.notna(sheet.loc[endog, i - 1]) and sheet.loc[endog, i - 1] != '':
            log_y_tm1 = np.log(sheet.loc[endog, i - 1])
        else:
            log_y_tm1 = 0

        # Forecast log(y_t), then convert back
        log_y_t = log_y_tm1 + dlog_y_t
        forecast = round(np.exp(log_y_t), 4)
        sheet.loc[endog, i] = forecast
def macro_log_forecast_mixed(endog, exo_vars, sheet, coef, start_period, end_period,lags_endog=None, exo_lags=None, diff_exo_vars=None, diff_dep=True):
    """
    Forecast with option to keep dependent variable in log-level (diff_dep=False) or difference it (diff_dep=True).
    Similar parameters to your original function.
    """
    if lags_endog is None:
        lags_endog = [1]
    if exo_lags is None:
        exo_lags = {var: [0] for var in exo_vars}
    if diff_exo_vars is None:
        diff_exo_vars = []

    global Assump, AB_Projections

    def get_val(var, t):
        if var in Assump and t in Assump[var]:
            return Assump[var][t]
        elif var in sheet.index and t in sheet.columns:
            return sheet.loc[var, t]
        elif sheet is CER_Projections and var in AB_Projections.index and t in AB_Projections.columns:
            return AB_Projections.loc[var, t]
        else:
            return None

    for i in range(start_period, end_period + 1):
        if i in sheet.columns and pd.notna(sheet.loc[endog, i]) and sheet.loc[endog, i] != '':
            continue

        X = [1]  # intercept

        # Exogenous variables
        for var in exo_vars:
            for lag in exo_lags.get(var, [0]):
                t = i - lag
                if var in diff_exo_vars:
                    t_minus_1 = t - 1
                    val_t = get_val(var, t)
                    val_tm1 = get_val(var, t_minus_1)
                    if val_t and val_tm1 and val_t > 0 and val_tm1 > 0:
                        dlog_val = np.log(val_t) - np.log(val_tm1)
                    else:
                        dlog_val = 0
                    X.append(dlog_val)
                else:
                    val = get_val(var, t)
                    log_val = np.log(val) if val and val > 0 else 0
                    X.append(log_val)

        # Lagged endogenous diffs only if diff_dep True
        if diff_dep:
            for lag in lags_endog:
                t_lag = i - lag
                if (t_lag in sheet.columns) and (t_lag - 1 in sheet.columns):
                    y_t = sheet.loc[endog, t_lag]
                    y_tm1 = sheet.loc[endog, t_lag - 1]
                    if pd.notna(y_t) and pd.notna(y_tm1) and y_t != '' and y_tm1 != '' and y_t > 0 and y_tm1 > 0:
                        dlog_y = np.log(y_t) - np.log(y_tm1)
                    else:
                        dlog_y = 0
                else:
                    dlog_y = 0
                X.append(dlog_y)

        # Calculate predicted value
        pred = sum(c * x for c, x in zip(coef, X))

        if diff_dep:
            # predicted diff(log(y))
            if (i - 1) in sheet.columns and pd.notna(sheet.loc[endog, i - 1]) and sheet.loc[endog, i - 1] != '' and sheet.loc[endog, i - 1] > 0:
                log_y_tm1 = np.log(sheet.loc[endog, i - 1])
            else:
                log_y_tm1 = 0
            log_y_t = log_y_tm1 + pred
        else:
            # predicted log(y_t) directly
            log_y_t = pred

        forecast = round(np.exp(log_y_t), 0)
        sheet.loc[endog, i] = forecast

#8.ALBERTA REGRESSION AND FORECASTS.
#8A.REGRESSION.
Cons_Dur_Param = make_regression('simple', 'Cons_Durables', ['Cons_Conf', 'Disposable_Income'])

NDURS_Param = make_regression('mixed', 'NDURS',['Cons_Conf', 'EXP_PR', 'WTI', 'Disposable_Income'],diff_exo_vars=['Cons_Conf', 
            'EXP_PR', 'WTI', 'Disposable_Income'],exo_lags={'EXP_PR': [1]})

ME_Param = make_regression('mixed', 'M&E',['WTI', 'Cons_Conf', 'EXP_PR'],diff_exo_vars=['WTI', 'Cons_Conf', 'EXP_PR'],
    exo_lags={'EXP_PR': [1]})

Res_Param = make_regression('mixed', 'Res_Investment',['Housing_Starts', '5-Year Bond'],diff_exo_vars=['Housing_Starts', '5-Year Bond'],
    exo_lags={'5-Year Bond': [1], 'CPI': [2]})

IPP_Param = make_regression('mixed', 'IPP', ['EXP_PR'],diff_exo_vars=['EXP_PR'],exo_lags={'EXP_PR': [1]})

NRes_Param = make_regression('mixed', 'Non_Res_Struct',['Capex', 'WTI', 'EXP_PR'],diff_exo_vars=['Capex', 'WTI', 'EXP_PR'],
    exo_lags={'EXP_PR': [1]})

Gov_Param = make_regression('mixed', 'Government',['Debt', 'WTI'],diff_exo_vars=['Debt', 'WTI'],exo_lags={'Debt': [1], 'WTI': [0]})

Export_Param = make_regression('mixed', 'Tot_Exports',['Exchange_Rate', 'USA_GDP', 'WTI', 'Tariff_Dummy','E_USA_TF'],diff_exo_vars=['Exchange_Rate', 'USA_GDP', 'WTI'],
    exo_lags={'Tariff_Dummy':[0],'WTI': [1]})

Import_Param = make_regression('mixed', 'Tot_Imports',['Cons_Conf', 'Exchange_Rate', 'WTI','Tariff_Dummy','E_CAN_TF'],
     diff_exo_vars=['Cons_Conf', 'WTI','Exchange_Rate', 'Tariff_Dummy'],exo_lags=None)

NPISH_Param = make_regression('simple', 'NPISH', [])

Source_Pop_Param = make_regression('simple', 'Source_Pop', [])

Inv_Param = make_regression('linear', 'Investment_Inventories', [])

LF_Param = make_regression(model_type='mixed',target='Labour_Force',exo_vars=['Pop', 'Per_Wage'],diff_exo_vars= ['Pop', 'Per_Wage'],lags=1,
           exo_lags=None)

Emp_Param = make_regression(model_type='mixed',target='Employment',exo_vars=['GDP', 'EXP_PR', 'WTI','Labour_Force'],diff_exo_vars=['GDP'
            , 'WTI','Labour_Force'],lags=0,exo_lags={'EXP_PR': [1],'WTI':[1]})

CPI_Param = make_regression(model_type='mixed',target='CPI',exo_vars=['GDP', 'EXP_PR', 'CAN_CPI','WTI'],
            diff_exo_vars=['GDP', 'EXP_PR','CAN_CPI','WTI'],lags=1,exo_lags={'EXP_PR': [0],'GDP':[1]})

H_Param = make_regression(model_type='mixed',target='Housing_Starts',exo_vars=['Pop','CPI','5-Year Bond'],
          diff_exo_vars=['Pop','CPI'],lags=0,exo_lags={'5-Year Bond': [1],'Pop':[0,1]})

#8B.GDP COMPONENT FORECASTS.
for i in range (start_period, end_period):
    macro_linear('Investment_Inventories',[],AB_Projections,Inv_Param,i,i+1,lags_endog=[1],lags_exo=None)
    macro_diff_log('NPISH',[],AB_Projections,NPISH_Param,i,i+1,lags_endog=[1],lags_exo=None)
    macro_diff_log('Source_Pop',[],AB_Projections,Source_Pop_Param,i,i+1,lags_endog=[1],lags_exo=None)
    macro_diff_log('Cons_Durables',['Cons_Conf','Disposable_Income'],AB_Projections,Cons_Dur_Param,i,i+1,lags_endog=[1],lags_exo=None)
    macro_diff_log_mixed('NDURS',['Cons_Conf','EXP_PR','WTI','Disposable_Income'],AB_Projections,NDURS_Param,i,i+1,lags_endog=[1],exo_lags={'EXP_PR'
    :[1]},diff_exo_vars=['Cons_Conf','EXP_PR','WTI','Disposable_Income'])
    macro_diff_log_mixed('M&E',['WTI','Cons_Conf','EXP_PR'],AB_Projections,ME_Param,i,i+1,lags_endog=[1],exo_lags={'EXP_PR'
    :[1]},diff_exo_vars=['WTI','Cons_Conf','EXP_PR'])
    macro_diff_log_mixed('Res_Investment',['Housing_Starts','5-Year Bond'],AB_Projections,Res_Param,i,i+1,lags_endog=[1],exo_lags={'5-Year Bond':[1],'CPI':[2]},
    diff_exo_vars=['Housing_Starts','5-Year Bond'])
    macro_diff_log_mixed('IPP',['EXP_PR'],AB_Projections,IPP_Param,i,i+1,lags_endog=[1],exo_lags={'EXP_PR':[1]},diff_exo_vars=['EXP_PR'])
    macro_diff_log_mixed('Non_Res_Struct',['Capex','WTI','EXP_PR'],AB_Projections,NRes_Param,i,i+1,lags_endog=[1],exo_lags={'EXP_PR':[1]},
    diff_exo_vars=['Capex','WTI','EXP_PR'])
    macro_diff_log_mixed('Tot_Exports',['Exchange_Rate', 'USA_GDP', 'WTI', 'Tariff_Dummy','E_USA_TF'],AB_Projections,Export_Param,i,i+1,lags_endog=[1],
    exo_lags={'Tariff_Dummy':[0],'WTI': [1]},diff_exo_vars=['Exchange_Rate', 'USA_GDP', 'WTI'])
    macro_diff_log_mixed('Tot_Imports',['Cons_Conf', 'Exchange_Rate', 'WTI','Tariff_Dummy','E_CAN_TF'],AB_Projections,Import_Param,i,i+1,lags_endog=[1],
    exo_lags=None,diff_exo_vars=['Cons_Conf', 'WTI','Exchange_Rate', 'Tariff_Dummy'])
    macro_diff_log_mixed('Government',['Debt','WTI'],AB_Projections,Gov_Param,i,i+1,lags_endog=[1],exo_lags={'Debt':[1],'WTI':[0]},
    diff_exo_vars=['Debt','WTI'])