With emissions of 2.

5 Gt CO2 in 2017, India ranked third globally, trailing only

China (9.8 Gt) and the US (5.3 Gt). Coal accounts for the bulk of India’s
contemporary primary energy supply, 58.1% in 20151, and is projected to continue
to play an important role indefinitely into the future, 42–50% by 20472. The share
of electricity in the overall energy system is predicted to rise from the current level
of 16 to 25–29% in 2047. In absolute terms, the demand for electricity is expected
to increase by as much as a factor of 4 over this time period.

The capacity for power generation in India amounted to 344 GW in 2018 of which
coal accounted for 197 GW (57%), hydro 49.8 GW (14%), wind 34.0 GW (10%),
gas 24.9 GW (7%), and solar 21.7 GW (6%) with the balance represented by a
combination of biomass 8.8 GW (3%) and nuclear 6.8 GW (2%).

The capacity factor (CF) is defined as the fraction of power generated by a

particular facility relative to its nameplate potential. Capacity factors for renewable
sources are typically much lower than those for coal, gas and nuclear plants given
the intermittent nature of the energy sources for the former. Renewables accounted
for <7.6% (1.3 PWh) of the total power consumed by India in 2018. NITI Aayog
set a target of 175 GW of renewable capacity for 2022, 160 GW of which would be
in the form of either wind or solar. Following these considerations, assessing
feasible renewable pathways to decarbonize India’s energy sector offers an
important and urgent challenge.

This paper considers the possibility of much higher levels of renewables for India
in the future. For present purposes, we refer to the combination of wind and solar
as renewables. There is a clear need for an integrated view of the potential for a
low-carbon future in India. This paper represents an integrated view of all
components of India’s electricity system and transmission to meet power demand
on hourly basis. It incorporates both thorough assessment of the potential for
renewables accounting at the same time for the practical operational limitations of
power systems. Detailed estimates for the physical (cost unconstrained) potentials
for wind (onshore and offshore) and solar PV are conducted. The overall objective
is to identify the least cost options to satisfy targets for incorporation of specific
levels of renewables in the overall power system. Five regional grids are
considered and the paper addresses requirements for power for each of these grids
on an hourly basis over a typical year.

investments in wind and solar could provide a cost competitive alternative to what
could otherwise develop as a coal dominated future for India’s power system while
contributing at the same time to a reduction of as much as 80% in emissions of
CO2.
LOGISTIC REGRESSION

Logistic regression is a workhorse in machine learning, particularly useful for

classification problems where the outcome variable can have two distinct
categories. Here's a breakdown of its key uses:

Predicting Binary Outcomes:

Logistic regression excels at predicting the probability of an event happening or

not happening. For instance:

Will a customer churn (cancel their subscription) or not?

Is an email spam or not?

Does a patient have a certain disease based on symptoms?

Classification Tasks:

By predicting probabilities, logistic regression can be used for classification tasks.

Imagine you want to classify a loan application as high-risk or low-risk. The model
would estimate the probability of defaulting on the loan, and based on a chosen
threshold (e.g., a probability of over 50% is high-risk), classify the application.

Understanding Relationships:

Logistic regression can reveal relationships between independent variables and the
binary outcome. The coefficients and odds ratios help understand how changes in
one variable affect the probability of the outcome.

Applications in Various Fields:

Logistic regression is widely used in finance, healthcare, marketing, and other
domains due to its ability to handle binary classification and provide interpretable
results.

There are two main ways to interpret the results of a logistic regression model:

Interpreting Coefficients:

These are the numbers associated with each independent variable in the model.
They tell you the direction and strength of the relationship between the variable
and the predicted outcome (binary).

Positive coefficient: As the value of the variable increases, the log odds of the
event occurring increases, leading to a higher probability of the event.

Negative coefficient: As the value of the variable increases, the log odds (and
probability) of the event occurring decrease.

However, coefficients are difficult to interpret in terms of magnitude. They

represent the change in the log-odds of the event, not the probability itself.

Interpreting Odds Ratios:

Odds ratios (Exp B in some outputs) are more intuitive for understanding the effect
of a variable. They represent the change in odds of the event happening for a one-
unit increase in the independent variable, holding all other variables constant.

Odds ratio > 1: Indicates that the odds of the event increase as the variable
increases.

Odds ratio < 1: Indicates that the odds of the event decrease as the variable
increases.
For example, an odds ratio of 2 for a certain variable means that a one-unit increase
in that variable makes the event twice as likely to occur.

Additional Interpretations:

Predicted Probabilities: Logistic regression doesn't directly give probabilities, but

some software (including Excel) might provide predicted probabilities for each
data point. These values range from 0 (impossible) to 1 (certain) and reflect the
model's prediction of the event for that specific point.

Model Fit Statistics: Look for metrics like Akaike Information Criterion (AIC) or
Schwarz's Bayesian Criterion (BIC). Lower values indicate a better fit for the
model.

Multiple linear regression

Use Multiple Linear Regression When:

 You are predicting a continuous outcome variable. This means the variable
can take on any value within a range. Examples include:

o Predicting house prices based on size, location, and number of

bedrooms.

o Forecasting sales figures based on marketing spend and economic

indicators.

o Estimating patient wait times based on arrival time and number of

patients waiting.

Use Logistic Regression When:

 You are predicting a binary outcome variable. This means the variable can
only have two distinct categories. Examples include:

o Classifying emails as spam or not spam.

 o Predicting customer churn (cancel subscription) or not churn.

o Diagnosing a disease based on symptoms (positive or negative).

Relationship between Independent and Dependent Variables:

Direction of impact: The signs of the regression coefficients (positive or negative)

indicate the direction of the relationship between each independent variable and the
dependent variable.

A positive coefficient suggests that as the independent variable increases, the

dependent variable tends to increase as well (and vice versa for negative
coefficients).

Strength of impact: The absolute value of the coefficient (ignoring the sign)
indicates the relative strength of the relationship. Larger coefficients imply a
stronger influence of the independent variable on the dependent variable. However,
the magnitude itself isn't always directly interpretable; consider standardized
coefficients for a more comparable measure.

Significance of the Relationship:

Statistical tests like p-values associated with each coefficient tell you whether the
observed relationship is likely due to chance or a genuine effect of the independent
variable on the dependent variable.

A low p-value (typically below 0.05) suggests the relationship is statistically

significant, meaning it's unlikely to be random.

Overall Model Fit:

R-squared (coefficient of determination) indicates the proportion of variance in the

dependent variable explained by the model. It ranges from 0 to 1, with higher
values suggesting a better fit (the model explains more of the variation). However,
R-squared doesn't necessarily imply causality.
Adjusted R-squared penalizes the model for adding more variables, providing a
more accurate measure of fit for models with many predictors.

Predictive Power:

The regression equation allows you to predict the dependent variable for new data
points with known values of the independent variables. However, these predictions
are estimates with some associated error.

ANNOVA

Sample means are same

Sample means are different

FACTOR ANALYSIS
KMO (Kaiser-Meyer-Olkin) and Bartlett's test for sphericity are two statistical tests
used together to assess the sampling adequacy for exploratory factor analysis
(EFA). Here's how they work:

1. Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy:

 This test measures the strength of the relationships between the variables
you're analyzing.
 KMO values range from 0 to 1, with higher values indicating better sampling
adequacy for EFA.
 Generally:
o KMO > 0.8: Very good
o KMO > 0.6: Acceptable
o KMO < 0.5: Not recommended for EFA (consider increasing sample
size or collecting more data)

2. Bartlett's Test of Sphericity:

 This test checks if the correlation matrix of your variables is spherical. A

spherical correlation matrix implies that there are no significant correlations
between the variables, which wouldn't be ideal for EFA (since EFA aims to
identify underlying factors explaining those correlations).
 Bartlett's test results in a p-value.
 You want a statistically significant p-value (typically less than 0.05) to reject
the null hypothesis of sphericity. This indicates that there are sufficient
correlations between the variables for EFA to be useful.

Interpretation:
  Ideally, you want a high KMO value (above 0.6) and a significant Bartlett's
test (p-value < 0.05). This suggests that your data has strong enough
relationships between variables for EFA to be appropriate.
 If either test fails to meet these criteria, it might be advisable to:
o Increase your sample size (if possible)
o Consider alternative data collection methods
o Explore alternative dimensionality reduction techniques that might be
less sensitive to these assumptions (e.g., Principal Component ana

COMMUNALITIES

 In multiple linear regression, R-squared represents the proportion of variance

in the dependent variable that can be explained by the independent
variables included in the model.

Communalities (Exploratory Factor Analysis):

 In exploratory factor analysis (EFA), communalities represent the proportion

of variance in each individual variable that can be explained by
the underlying common factors extracted by the analysis.
 They also range from 0 to 1, with higher values indicating that a larger share
of the variable's variance is explained by the common factors.
 Communalities reflect how well each variable is represented by the
common factors. A low communality might suggest the variable is not well-
suited for the current factor structure or may require additional factors to
explain its variance.
  An eigenvalue represents the proportion of variance explained by the
corresponding eigenvector (direction) in the data.
 Eigenvalues are typically arranged in descending order, with the first
eigenvalue explaining the most variance, the second explaining the second-
most variance, and so on.

Using Eigenvalues in EFA:

 By looking at the distribution of eigenvalues, you can gain insights into the
number of factors to retain in your EFA model.
o A common rule of thumb is to keep factors with eigenvalues greater
than 1. This suggests they explain at least as much variance as a
single original variable.
o The more eigenvalues exceeding 1, the more complex the underlying
structure in your data, potentially involving multiple important factors.