If we don't know your hypothesis, then we can't help. 

Statistical tests are neither correct nor incorrect. They either test the hypothesis of interest or they don't. And among those that do, they are either the most efficient way to test it, or they aren't. 

So tell us more. What is your study hypothesis and what does it predict about the differences between the groups?

from the original poster:

Dear Sir,

I have conducted experiments to Check the "Effects of larval hosts on the biological traits of mosquito"

Hypothesis:

Different larval hosts alters the biological traits of mosquitoes.

Brief description:

The data of larval weights,pupal weights,larval duration,pupal duration, adult emergence, longevity of adults,fecundity/female, hatching% were taken.Data of all parameters for each host was analysed by ANOVA(CRD) and means were compared by LSD-test to check significant difference.

Sir, I also have applied Turkey test but it shows similar lettering for most of the values irregardless of their numerical magnitude i.e,28 and 158 both have same letter A.Therefore,I then used LSD-test that shows different lettering for both of these and all such cases.

Meanwhile,the question arise can i use LSD-test instead of Turkey test? because results of comparison LSD test shows are more suitable to my hypothesis.

Regards

Rizwan Mustafa

Dear Rizwan

I would recommend that you read up on post-hoc tests. There is an excellent worked and commented example here: http://pages.uoregon.edu/stevensj/posthoc.pdf

Basically, Fisher's original post-hoc test, the LSD, has the problem that it produces inaccurate p-values by performing multiple t-tests on the same sample. This leads to a situation where the false-discovery rate increases with each pair of means that is compared. 

With three groups, we have three possible two-way comparisons, giving us a false-discovery rate of 14% (recall that the false-discovery rate should be 5%). With four groups, the false-discovery rate is 26%, and with five groups it's 40%.

So the LSD is fatally flawed, and should not be used at all. 

Tukey (not Turkey!) produced a post-hoc test that overcomes this problem so long as the comparisons are only of individual groups (so not comparisons involving two groups versus a third group). The reason why your comparisons are significant with the LSD and not significant with Tukey is because the false discovery rate for the LSD is unacceptable. You are better off with the Tukey. 

You can read about the LSD here, at GraphPad's excellent website: http://www.graphpad.com/guides/prism/6/statistics/index.htm?stat_fishers_lsd.htm

@Ronán, thank you very much sir for providing such value able information.I got it now. 

If I were you I will use contrast analysis if there is any prior information about the structure of treatments is available. Otherwise, I will use Tukey HSD test beyond ANOVA.

You do not state whether your initial ANOVA showed any treatment effect. If the overall ANOVA is negative then post hoc testing you describe is not relevant. 

can we use tukey's test for 40 treatments?

LSD means least significant differences. That is, LSD test calculates the minimum difference between two treatments medias, of a k treatments completely random experimental design (analyzed by 1W ANOVA), necessary for to declare these two treatments significant. This is because the p real (p_R) is greater than the LSD p (p_LSD). As example if you conduct LSD with p_LSD=0.05 then p_R>0.05 . If your LSD test declare non-significant two treatments you can be sure of conclusion; however if your LSD test declare significant two treatments it is possible that these two treatment have not significant difference. Therefore LSD test is not suitable for any 1W ANOVA with k>2. A more suitable test is Tukey.

what about multible T tests ? Can we use it for geting the significance among many treatments in staid of LSD?

LSD is not so reliable when you want to compare between more than three groups (treatments). In this case, it is safer to use HSD (Tukey) as it takes the number of groups into account.

To my understanding, your question revolve around protection against TYPE I error and LSD is poor against that compared to the means separation technics. And the more the number of treatments the higher the chance of commiting type I error, that is why for treatments >4 is better to use other means separation test.

Ali Elhakeem you nailed it. I had same challenge but your recommendation worked perfectly. Thank you.

LSD is one of the Post-Hoc test that can be applied if the initial ANOVA test is significant @ p