Pages

Featured Posts

Friday, February 17, 2012

Null hypothesis for Hosmer and Lemeshow goodness of fit test

In logistic regression, we find a goodness of fit statistic with a p value displayed alongwith it. The null hypothesis is that the model is fit. If the p value is less than 0.05 and the null hypothesis is rejected, it means that the model is not fit.

Sunday, December 18, 2011

Difference between error and residual ...

Error is the difference between the observed value in a sample/subject and the true value in the population (which is actually not known).
whereas Residual is the difference between the observed value and the predicted (or estimated value) from our regression equations.
So, even they may sound quite similar but are actually quite different. In regression, we have to be very careful about the residual diagnostics. They are very vital.

Monday, December 5, 2011

Maternal mortality rate .... Is it the same as Maternal mortality ratio ?

No, Maternal mortality rate is the number of maternal deaths per 1000 women in the reproductive age group. Whereas Maternal mortality ratio is the number of maternal deaths per 1 lac live births. These two are different measures.
Source: link is http://www.who.int/making_pregnancy_safer/documents/measuring_maternal_mortality.pdf

Saturday, December 3, 2011

p-value and confidence interval

p - value is used in hypothesis testing. It reveals the probability of 'chance' to be responsible for the difference. The less the p value the less the probability that the difference is due to chance. Conventionally, a cut off of 5% or 0.05 or one in twenty is considered for saying that there is a statistically significant difference. But, always keep in mind that the difference in 0.049 and 0.051 is not great. The probability that the difference is due to chance is 4.9% and 5.1% respectively in the example given in the earlier line. So they are telling nearly the same story. But going by our cutoff of 5%, one is statistically significant and the other is statistically not significant. Human judgement, common sense and experience should prevail for interpretation of the result. One more word of caution: Increasing the sample size tends to magnify the differences and chances for finding a statistically significant difference increases.

95% Confidence interval arrived from our sample is the probability that the population value lies in that interval 95% of the time. The condition is that the sample should be representative of the population.