Fix typos in documentation.

docs/Intro to lifelines.rst

@@ -522,7 +522,7 @@ Smoothing the hazard curve
 
 Interpretation of the cumulative hazard function can be difficult -- it
 is not how we usually interpret functions. (On the other hand, most
-survival analysis is done using the cumulative hazard fuction, so understanding
+survival analysis is done using the cumulative hazard function, so understanding
 it is recommended).
 
 Alternatively, we can derive the more-interpretable hazard curve, but
@@ -544,7 +544,7 @@ intervals, similar to the traditional ``plot`` functionality.
     ax = naf.plot_hazard(bandwidth=b)
     naf.fit(T[~dem], event_observed=C[~dem], label="Non-democratic Regimes")
     naf.plot_hazard(ax=ax, bandwidth=b)
-    plt.title("Hazard function of different global regimes | bandwith=%.1f"%b);
+    plt.title("Hazard function of different global regimes | bandwidth=%.1f"%b);
     plt.ylim(0,0.4)
     plt.xlim(0,25);
 
@@ -555,8 +555,8 @@ intervals, similar to the traditional ``plot`` functionality.
 It is more clear here which group has the higher hazard, and like
 hypothesized above, both hazard rates are close to being constant.
 
-There is no obvious way to choose a bandwith, and different
-bandwidth can produce different inferences, so best to be very careful
+There is no obvious way to choose a bandwidth, and different
+bandwidths can produce different inferences, so best to be very careful
 here. (My advice: stick with the cumulative hazard function.)
 
 .. code:: python
@@ -566,7 +566,7 @@ here. (My advice: stick with the cumulative hazard function.)
     ax = naf.plot_hazard(bandwidth=b)
     naf.fit(T[~dem], event_observed=C[~dem], label="Non-democratic Regimes")
     naf.plot_hazard(ax=ax, bandwidth=b)
-    plt.title("Hazard function of different global regimes | bandwith=%.1f"%b);
+    plt.title("Hazard function of different global regimes | bandwidth=%.1f"%b);
 
 
 
@@ -607,7 +607,7 @@ of time to birth. This is available as the ``cumulative_density_`` property afte
 
 .. code:: python
 
-    kmf.cumulative_density_
+    print kmf.cumulative_density_
     kmf.plot() #will plot the CDF
 
 
@@ -617,7 +617,7 @@ Left Truncated Data
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 Another form of bias that can be introduced into a dataset is called left-truncation. (Also a form of censorship). 
-This occurs when individuals may die even before ever entrying into the study. Both  ``KaplanMeierFitter`` and ``NelsonAalenFitter`` have an optional arugment for ``entry``, which is an array of equal size to the duration array.
+This occurs when individuals may die even before ever entering into the study. Both  ``KaplanMeierFitter`` and ``NelsonAalenFitter`` have an optional arugment for ``entry``, which is an array of equal size to the duration array.
 It describes the offset from birth to entering the study. This is also useful when subjects enter the study at different
 points in their lifetime. For example, if you are measuring time to death of prisoners in 
 prison, the prisoners will enter the study at different ages. 

docs/Survival Analysis intro.rst

@@ -112,11 +112,11 @@ information at :math:`t=10`).
 Survival analysis was originally developed to solve this type of
 problem, that is, to deal with estimation when our data is
 right-censored. Even in the case where all events have been
-observed, i.e. no censorship, survival analysis is still a very useful
+observed, i.e. no censorship, survival analysis is still a very useful tool
 to understand durations.
 
 The observations need not always start at zero, either. This was done
-only for understanding in the above example. Consider the example of
+only for understanding in the above example. Consider the example where
 a customer entering a store is a birth: a customer can enter at
 any time, and not necessarily at time zero. In survival analysis, durations
 are relative: individuals may start at different times. (We actually only need the *duration* of the observation, and not

docs/Survival Regression.rst

@@ -41,9 +41,9 @@ variables ``un_continent_name`` (eg: Asia, North America,...), the
 ``regime`` type (eg: monarchy, civilan,...) and the year the regime
 started in, ``start_year``.
 
-Aalens additive model typically does not estimate the individual
+Aalen's additive model typically does not estimate the individual
 :math:`b_i(t)` but instead estimates :math:`\int_0^t b_i(s) \; ds`
-(similar to estimate of the hazard rate using ``NelsonAalenFitter``
+(similar to the estimate of the hazard rate using ``NelsonAalenFitter``
 above). This is important to keep in mind when analzying the output.
 
 .. code:: python
@@ -156,7 +156,7 @@ above). This is important to keep in mind when analzying the output.
 
 
 
-I'm using the lovely library ``patsy`` <https://github.com/pydata/patsy>`__ here to create a
+I'm using the lovely library `patsy <https://github.com/pydata/patsy>`__ here to create a
 covariance matrix from my original dataframe.
 
 .. code:: python
@@ -391,9 +391,9 @@ Prime Minister Stephen Harper.
 Cox's Proportional Hazard model
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-New in 0.4.0 is the implementation of the Propotional Hazard's regression model (implemented in 
-R under ``coxph``). It has a similar API to Aalen's Additive model. Like R, it has a ``print_summary``
-function that prints a tabuluar view of coefficients and related stats. 
+New in 0.4.0 is the implementation of the Cox propotional hazards regression model (implemented in 
+R under ``coxph``). It has a similar API to Aalen's additive model. Like R, it has a ``print_summary``
+function that prints a tabular view of coefficients and related stats. 
 
 This example data is from the paper `here <http://socserv.socsci.mcmaster.ca/jfox/Books/Companion/appendix/Appendix-Cox-Regression.pdf>`_.
 
@@ -459,7 +459,7 @@ Model Selection in Survival Regression
 With censorship, it's not correct to use a loss function like mean-squared-error or 
 mean-absolute-loss. Instead, one measure is the c-index, or concordance-index. This measure
 evaluates the ordering of predicted times: how correct is the ordering? It is infact a generalization
-of AUC, another common loss function, and is interpretted similarly: 
+of AUC, another common loss function, and is interpreted similarly: 
 
 * 0.5 is the expected result from random predictions,
 * 1.0 is perfect concordance and,